Professional Documents
Culture Documents
<X>^O<KXX><XXXXKXXXXX><XXX><><X><XXXHX>O<X:K><) -^
UNIVERSITY OF CALIFORNIA
AT LOS ANGELES
A TREATISE
ADJUSTMENT OF OBSERVATIONS,
APPLICATIONS TO GEODETIC
OF PRECISION.
W. WRIGHT,
T.
B.A.,
Civil Engineer,
O Messkunst Zaum
Wtr dir
der Phantasie,
will folgeit irrct nie.
HALI.ER.
NEW YORK
D.
23
VAN NOSTRAND,
27
WARREN
STREET.
Copyright,
1884.
T.
W. WRIGHT.
PRINTED BY
H.
27
J.
HEWITT,
NEW YORK.
ROSE STREET,
Q.
PREFACE.
especially with
reference
to
geodetic and
The
not noticed.
once assent
to.
As regards
notation
331
211005
PREFACE.
work
have aimed at
where descriptions
of instruments
can be found.
Special attention has been given to the explanation of
checks of computation, of approximate methods of adjustment, and of approximate methods of finding the precision
of the adjusted values.
But
in
order to see
how
far
it is
time
in finding the
found,
I
ities
of
it
in
many
cases tells so
it
has been
little.
have been careful to give references to original authoras far as I could ascertain them, and also to give lists
memoirs on
which
will be of use to
any
one desiring to follow those subjects farther. Of recent
I
writers I am indebted chiefly to Helmert and Zacharise.
special subjects
desire also to
me
advice.
TABLE OF CONTENTS.
CHAPTER
I.
Introduction.
..........
.........
.........
..........
...........
..........
............
The instrument,
External conditions,
The observer,
Theory
of probability,
Definite integrals.
Taylor's theorem
Examples
18
19
21
Interpolation,
25
Periodic series,
27
28
Notation,
CHAPTER
T^i?
The
arithmetic
mean
II.
Law of Error.
The
arithmetic
When
mean
the
....
...
...
.....
....
.....
.......
....
.....
Law
Law
29
31
32
33
34
37
38
42
44
44
44
........
..........
.........
..........
....
.........
....
.........
........
The average
error,
to
46
47
48
54
OD
CHAPTER
Adjtistinent
PAGE
57
59
65
66
67
III.
....
......
........
........
.......
.....
......
..........
......
........
.........
......
:
mean
69
71
Bessel's formula,
73
Peters' formula,
76
Approximate formulas,
The law of error tested by experience,
79
82
Caution as to the
Constant error,
89
tests of precision,
85
gi
Combining weights,
........
.........
.....
........
......
.....
.....
..........
.....
92
94
95
96
96
99
value,
Precision of a linear function of independently observed values,
Miscellaneous examples,
NOTE
I.
On
An
approximate method,
Weighting when constant error is present,
Assignment of weight arbitrarily,
Combination of good and inferior work,
The weight a function of our knowledge,
General remarks,
NOTE
II.
On
.'
102
105
no
118
120
121
126
127
128
130
131
TABLE OF CONTENTS.
CHAPTER
IV.
Adjustment of Indirect
Oi>ser;>atioiis.
......
......
........
Forms
Forms
of solution
......
.......
.....
.....
.....
Combination
of indirect elimination,
of the direct
....
....
......
Time
.....
.....
.....
.....
......
..........
......
adjusted value,
Examples and
145
150
152
154
156
158
160
163
165
167
172
174
177
178
180
184
186
187
193
its
artifices of elimination,
CHAPTER
139
151
The method
PAGE
198
201
V.
General statement,
..........
......
.......
.........
........
......
Weight
of the function,
Programme
of solution,
213
214
217
224
224
227
238
242
243
247
CHAPTER
VI.
Triangulation.
PAGE
.......
.........
........
......
........
......
.......
........
......
....
General statement,
The method of independent angles,
The local adjustment,
.
Number
of local equations,
Number
The
250
252
255
258
259
259
261
of angle equations,
side equation's,
to the linear form,
263
266
Check computation,
268
Position of pole,
Number of side equations,
271
Reduction
of a quadrilateral
272
273
273
.....
........
.....
........
.....
.....
......
........
.......
.......
280
282
Solution by correlates,
Precision of the adjusted values,
Solution in two groups,
284
286
293
288
298
of a triangulation net,
Artifices of solution,
Adjustment
300
302
....
.........
......
........
......
..........
.......
302
303
3*3
precision,
315
316
32
3 22
On
Discrepancy
Disci epancy
Discrepancy
in
32^
Adjustment
323
azimuth,
and longitude,
33 6 339
>
34 1
34 2
343
in bases,
in latitude
33
333.336
.....
347
347
TABLE OF CONTENTS.
CHAPTER
to
Application
VII.
Base-Line Measurements.
PAGE
......
General statement
Precision of a base-line measurement,
353
356
....
in a triangulation,
Approximate solutions
356
360
CHAPTER
is
considered,
362
363
.....
.....
to
365
368
VIII.
...........
........
.......
.....
........
.......
Appliiation
Spirit levelling,
..........
Rigorous solution,
When
When
349
350
Levelling.
371
374
377
Trigonometrical levelling,
To
To
To
mean
coefficient of refraction,
.....
....
....
.....
Approximate methods
of adjustment,
CHAPTER
to
Application
and
to
Calibration
...........
........
Line measures,
Calibration of thermometers,
Application
General statement,
Applications,
Periodic phenomena,
to
TABLES,
I.
II.
386
388
388
392
of
395
401
Applications,
APPENDIX
APPENDIX
386
X.
..........
..........
.
384
IX.
CHAPTER
378
382
Historical note,
The law of error,
408
413
417
421
....
....
427
429
435
I.
INTRODUCTION.
THE
measurement of a quantity are, the observer, the instrument employed, and the conditions under which the measurement is made.
If the measure of a quantity is
i. The Instrument.
factors that enter into the
value.
The many
external influences at
is
of
work hinder
result,
it is
measured.
in
showing
in the
This opened
12
the
way
servation.
circles
finally in the
reading microscopes
now almost
universally
The culminating
it is
ures
should suspect the chain to be in error and proceed to examine it before further measuring. So discrepancies found
in measurements made with the same measure at different
temperatures have shown the necessity of finding the length
of the measure at some fixed temperature, and then applying a correction for the length at the temperature at which
the measurement is made.
Corrections to directly measured values are thus seen to
be necessary, and to be due to both internal and external
causes.
The internal causes arising from the construction
of the instrument are seen to be in great measure capable
of elimination.
From geometrical
server can
tell
The instrument-maker
cannot,
and more
INTRODUCTION.
13
when
circle
is
influence of the periodicity by employing a number of verniers or microscopes placed at equal intervals around the
circle.
It
happens that
this
still
is
more
An
aimed
at.
is
14
doned on
perature.*
The
may sometimes be
by the method of observation.
American Journal of
INTRODUCTION.
In the
ment
is
15
measurement of horizontal angles where the instruplaced on a lofty station, the influence of the sun
signals observed
in
Atmospheric refraction
If
the correction
itself is
small this
is
allowable.
As
is
a variable factor
all
new
The
should not begin work unless he considers that he himnormal condition. If he is not in that condition
self is in his
16
he introduces an unknown disturbing element unnecessaHe is also more liable to make mistakes in his readFor the same reason he should not
ings and in his record.
continue a series of observations too long at one time, as
from fatigue the latter part of his work will not compare
In time-determinations, for infavorably with the first.
stance, nothing is gained by observing from dark until
rily.
daylight.
centre.
observers have,
is
known
Thus
servation.
is
of the work.
in
INTRODUCTION.
It
is
I/
the same
We
When
known
all
ceding pages.
in the con-
ditions of sufficient importance to have attracted the observer's attention when making the observations, but he
his
this.
With our
limited
the
As soon
errors.
Having, therefore, taken all possible precautions in making the observations and applied all known corrections to
the observed values, the resulting values, which we shall in
future refer to as the observed values, may be assumed to
We
How
brought face
are, then,
Probability.
an event
is
number of
(i)
whose numerator
is
and
number
the
its
happening
is tt
-r,,
and of
its failing
dr
*"j
certainty
INTRODUCTION.
19
Thus
happen
is
8-^6
9
6.
Definite Integrals.
Let
ing,
=:
A'-sr
(a)
then regarding
To
find
the
value
of
as a constant in integrat-
we have
~f*dt
r~f**xdz
\J
Multiply each
/e^dt,
^/
y^.*
(b)
e~*dx,
which
g+dt\
)
f**>
I ds
v/
Jf
e^dt
To
yC
is
equal to
Hence
member by /
J
/-.
<r**+*
same
then
20
Integrating by parts,
aff*dt
we have
W
r*
L_
^J
+/
tra
/
= (-*!-}
fs>-aW\
2a /-
2 a*
^
d(af)
'
20*
Also, similarly,
(c)
To
The value
lows
may
be found approximately as
fol-
e'*
Expanding
in a series
we have
separately,
dt
a*
a
---= a --3
1.2 5
This series
is
convergent for
all
vergence
is
Integrating by parts,
Jfe-*dt=Jf-^-de^
2t
= -- a -i
2t
Hence
^
a
~~
_*<2
/V
,.
/
T dt
2.J?
INTRODUCTION.
21
But
S**
/a
e-*dt -- I
VTT
'~~J
.
f^*
e^dt
e~*dt
*J
r"
dt
finally,
" dt
~?
'
^T
~~
i + ^y ~ Sf +
/(Le'^dt
computed from the above formulas
value of
for
any
may be
numerical
a.
then
(O
By
/(
+ A) = / W +
/<
by putting
22
Let
(b)
F,
F=f(X,
and
let
and
X',
Y',
x, y,
F,
the corrections to
values, so that
XX'+x
Y=Y'+y
then
6'F
d'F
,,
d'F
'
'
where
6X'
6 Y'
and
f (X, Y, ...),... with respect to X, F,
for X, F.
then substituting X', Y',
If the corrections x, y,
are so small that their
and
be
neglected and they are
squares
higher powers may
F
dX
dY'
and
written
f (X', Y',
.) is written
dF, then
ating
which
is
ties.
We
shall use
venient.
Ex.
log
i.
(N+v)
If
-v
is
number N, required
to
express
INTRODUCTION.
We
23
have
\og(W+v)
^
=log^+ dN
(N+ v) = log N +
5N v
where <5 N is the tabular difference corresponding to one unit for the number.
For small numbers, however, it is better to take <5 N from the table for the
form
Thus from
the table
+ ^) = 3.8157902 + 0.0000664 v
+ v) 2.8157902 + 0.0006637 v
log (65.432 + ^) = 1.8157902 + 0.0066378 v
log (6543.2
log (654.32
Ex, 2. In a ten-place log. table where angles are given at regular intervals,
required to find log sin (A+a) when A is given in the table and a is a number of seconds less than the tabular interval.
We
have
log sin (A
+a)=\og
sin
=log
sin
A +-T-J
(A +
log sin
A + mod.
sin
i"
a cot
(A
Now,
log
mod.
log sin
7
log io
i"
=:
9.6377843
= 4.6855749
= 7.
10
10
1.3233592
log sin (A
= log
i") in
Hence
+ a)
sin ,4+log-'
i.3233592
+ logrt+ (log
cot
A+
diff. for
i"j
j-
Lipsiae, 1794.
24
Thesaurus,
amounts
"
the uncertainty of the last figure
the above process is in general sufficient when
Hence
to 4 units/'*
TT
dA.
10
1
(
so that
log sin (A
+a)=log
sin
A + 6A
a.
then
log sin (A +v)
=log
sin (A i
=log
sin
log sin
+ v)
AI + mod.
AI +
log"
log-
Expand log
sin
sin (68
i"
a cot
AI +
a\
-
1.3233592
log a
1.3233592
(log cot
16'
+ mod.
(log cot
A +
i
AI +-
X diff.
for
diff. for
i")|
a)
i")
3:
68
A!
a
16' 30"
2". 076
1.3233592
127
0.9237245
-?Xdiff.
i"
8,3893
+64
0.9237309
0.3172273
logtf
I7,4l6
1.2409582
log sin 68
Hence log
when
16' 30"
sin (68
the difference
is
9.9680022,271
16' 32" .076
expressed
+ v)=
in
9.9680039,687
8,3893 v
unit.
* Bremiker's edition of
Vega, translated
by Fischer.
Preface, p. 10.
INTRODUCTION.
With
it
will
be
log sin (A
=^o sn
when 6\
8.
2'
Interpolation.
16' 32"
So
i" for
It
far as interpolation is
concerned,
given
Function.
26
sin
A when A
is
any angle.
Let
= the
given angle.
d^ d^=i the first and second tabular differences.
then
log sin
Writing
= log sin
f^,-j
= logsin
AA-a
---
10
ioj
+ tfaho
a)-*-.
100
(i)
log sin
,4
= log sin
,4
,+
-f
-^\
(5
(2)
difference multiplied
the tabular log. sine.
To
From
16'
32".O76
the table,
log sin 68
i6'3o"=
di=
d%=
Hence, from equation
9.9680022,271
83,889 for 10" in units of the seventh decimal place
0,012
(i),
=2.076X8, 3889 +
^^
2
17,416
.'.
log sin 68
16' 32".
076
9.9680039,687
(10
2.076)
100
INTRODUCTION.
The
We
difference at
want
35"=
8,3889.
middle
it
2J
of the interval.
diff.
= 0.0005
= 8,3889
Diff.
required
And
2.076
0,00012
8,3894
17,416 as before.
8,3894
9.
Sin
where
lows
#-|-sin 20-\-
(?-(-sin
8=~
(;/
-|-sin
\)0
1)6
we may proceed
n being an integer,
as fol-
respectively.
If we divide the circumference of a circle into n equal
then each angle at the centre
parts at the points A, B,
.
^60
is ^
or
6,
A we
lines
find the
sum
to be also zero.
where
assumes
Hence
all
follows that
sin
in0 cos
i0
wti=
l/
2
sin in
I.
2^
v
h/2- cos
values from o to n
it
_v cos
- sn
'
;//
u
20
sin
m0
n
2
~>
/2- cos
l
///
20 -
28
braic
a,2 ,
sum
is
are quantities of the same kind, their alge[a], and the sum of their squares by
denoted by
[aa\ or [a*]
= a? + a* +
...
+ a\
Also,
ab~
= ab--a-j_
We
c,
of quantities a
all
_i_
**
an bn
symbol
[a to
cn
c^
CHAPTER
II.
Mean.
Arithmetic
M M
tity to
lt
T-M=A
where
J,,
Av
An
T and
We
development.
be
some
two
central values
if
the
greater than
change
in
it
as
it
3O
would
the
in
discrepant.
On the other hand, the taking of the central value is objectionable, because it gives the preference to a single one
of the observed values, while if these values are supposed
to be equally worth)' of confidence, as
all
reasonable to
it is
to the contrary,
knowledge
when
tion that
another
it
T=f(M
f(M,M,
M=
M.
tion, then necessarily f(M)
Let, then, V be a symmetrical function of
M=
M^
n>
and put
= f(V,V,.
where
.
//,
K,
powers
vol.
1876.
.)-\v\
* See
Reuschle, Crelle
tute
Insti-
Lombards, 1868; Astron. Nachr., 2068,2097; Stone, Month. Not. Roy. Astron. Soc.,
Astron. Nachr., 2092; Ferrero, Expos, del Met. dei Min. Quacfr., Florence,
xxxiii.
;
kl.
Abweichungssumme,
Leipzig, 1874.
31
The simplest symmetrical function of the observed values that can be chosen as the form for Fis their arithmetic
mean
that
'-
is,
If
^.
we
V equal
take
from equation
(i)
first,
equation
M M
1}
f(M
M M
whole)
is
the arithmetic
By adding
we have
12.
mean
equations
last
* Gauss'
sec. 3.)
term of
words
"
are,
Art.
(i),
V
The
..
\^
this equation
'
n
will
solet hypothesis."
(Theoria Motus,
lib.
if,
z,
32
Now,
after
if,
sum
[J]
of the
errors remains
for observing,
it is
may, when n
a very large
is
number, consider
We
LJ
to be an
may, therefore,
in this
case put
V= T
that
is,
arithmetic
is
From
we may
write our
V-M =v
n
where v^
tion,
t/ a ,
By
(a)
and
33
addition
nV -[M~\ = [v]
[>]=0.
that
is,
sum of the
the
residuals
residuals
is
is
equal
zero
(2)
;
to the
in
sum of
the negative
residuals.
V'
T''
(3)
From
equations
[vv]
f
[v'v
Hence by
and
(i)
(3),
a simple reduction,
[t^=[W] + *(r'
Now, |F'
-t
")
is
always
positive.
.'.
[W]>
[>'?']
given by Legendre.
Let us recapitulate the three forms of solution proposed
34
the n equations
Find
(1)
values of
all
possible values of
Fare M^
Mv
Fand
The
(2)
[v~\
is
!=
This gives
O.
or
n
the arithmetic mean.
(3) Solve simultaneously, making [w]
This gives
(V-M$+(V-M$+
(V- Mn
=a
)*
minimum.
= a min.
or
n
the arithmetic mean.
14. When the quantity measured is a function
of the quantities to be found. We pass now to the
more general but equally common case in which the ob-
are to be found.
Tand
the
35
let
unknowns X,
Y,
T=f(X,
F,
be
.
(i)
M M
(2)
in
vv
which a
.
ber of unknowns,
the values of X, F,
may be found
the
in the
if
and
substituted
by
ordinary algebraic methods,
will
if
the
number
of
But
equations
satisfy them exactly.
equations exceeds the number of unknowns, the values
found from a sufficient number of the equations will not
.
,-,
in
We
far
For simplicity
where a lt b^
6
in
Fare
to be found.
36
Find
(i)
all
possible values of
and
Y,
and combine
them.
whence
at
sets of
two equations
once
(a
A - a A} X=
Y
M,
0,M a
b,
(afa
,,)
(a,d 3
a 3 d^
(ajb t
a&)
Y= a,M
(aj),
a,b,)
X= b M,
(a&
a&)
X= b M^
t
a,M
- b,M,
a,M,
b^M
a M,
3
b,M,
a3 t
Y we
are
met by
the arithmetic
mean gives
all
equal quality.
It is
X and
which
is
infinite
number
therefore, insufficient.
(3)
The simultaneous
solution by
minimum.
My +
...
We
...
have
a rain.
we
37
as independent vari-
find
\_aa\X-\-\ab\
[ad]
Y=[aM~\
X+ [bb\ Y= [bM]
where
=
[ab] =
[aa\
a, a, -f- aji,
ajbi -j-
-f a 3 a 3
a& + aj,
mean of the
distances,
minimum.
weights is a minimum.* On this principle Legendre founded the rule of minimum squares, and he employed the rule
*
21101)5
38
The
Law
sum
residuals a
The
minimum.
most numerous
in the
If,
Hence
neighborhood of zero.
may
be as-
A be denoted by
J and A-\-dA is
tp(A)dd
suppose.
(i)
39
tending from b to
a,
and
is
by
A
Hence
it
(2)
is
(3)
The
determine
to
Among
where h
is
a constant.
This was, indeed, that selected by Laplace in his investigation Determiner le milieu que I' on doit pr entire entre f rot's
observations donnees d'un meme phtnomene (Mem. Acad. Paris,
From this form may be readily derived the results
1774).
stated in Art.
n.
The form
For
ities.
if
abilities of the
^>(4)
dJv
9>(4i) ^4i>
occurrence of errors
in
tween
$0(4) ^(4,)
Denote
Now,
.
this expression
by
</?,
(4)
so that
Jw are unknown.
,
J,,
</>
40
hypothesis.
we must have
Now, since log tp varies as
mum, and therefore by differentiation
log
</>,
dT
b log
'
dT
<>4
or
ffrL
i
*y^
IU ^
^i
.^
^4
n^;
A
/f
(i),
Art.
j3
"
dT
6 log
lu s y>(4
f\
</>
a maxi-
/g\
.
,
11,
^4
~dT
^/4
dT~dT
when
the
number
4 + 4+.
+4-0.
(7)
6 log
_ 6 log
p(4)
4 dJ
4 d4
When
=2
it
reduces to an identity.
k suppose.
4!
since
Now,
But when
72
then
<p
is
</>
is
to be a
maximum
= nk$.
maximum,
it
must be negative.
Hence, since
of logarithms
</>
equal to
y( J) =
positive, k
is
_
ce
o,
must be
we have
we have
dA =
/+
a
found from
/+
is
-00
6)
H V2K
and the law of error may be written
i
ft
or by putting
//"
*
\2
42
When this latter form is used it is only for greater convenience in writing, and h is to be looked on as a mere symbol standing for
As regards
J"
quantity
//",
A"
it is
to be a possible
Hence a
is
I?
being
of error
-j-
a and
a will be
a to
<p(A) its
this
expression
value,
/"+0
J~
which (see Art.
6)
reduces to
//.
the
sum
series.
It
fined the
symbols
and
//
in the
s. e.
Having de-
is
maximum when
evidently a
43
11
[J
] is
minimum
we assume
that
same quality, (i is
and the most probable value of the unknown would be found from the maxidifferent for the different observations,
mum
value of
pf-|
-~
I
Thus
J.
quantity
is
f2 1
L
^J
is,
if each of a large
that
quantity
is
differ
4 =/'(*.*
O-^i
4 =/*(*,* ...)-^
-
Hence
since
J,,
variables x, y,
J
.
a,
.
we must have
44
many
values of different
(b)
Since
r is
unit of measure in
it
Law
The
quantities,
MM
n;
that
is,
when
and
4, in
M,
is
M^
45
and an error Ja in
Now, an error 4 in
error J in F, according- to the relation
M^ produce an
= *,4+,4
J
and
this relation
can always be
(2)
satisfied
by combining any
oo to -f-co.
ranging from
The probability, therefore, of an error J in F may be written
value of 4, with
<p(J)
all
dJ
values of
A <U f 'Y*'*' - W
=H
(2),
and since J2
independent of
is
af
<W,
-w
*s
7T
But from
J,
J,,
^Js
Hence
= h^d_J
7T
<7 2
/
^/ -w
^-^v-vC-^OV
J --*w+w a s*
_'tA //
e
/,
/,
which
is
>- + =o
./
v^
of the form
That
is,
the
of the independently
that
is,
from
theorem
is
46
s. e.
mean
of the arithmetic
of
We have
.
Let
//
>u
Then
/*
That
is,
the
m.
s. e.
=:
m.
m.
-5 (ju-
+Afn )
mean
s. e.
of the arithmetic
s. e.
/<
+ ...
to
n terms)
observations is
Vn
part of that
of a single observation.
On
the
20.
1
6 that the m.
s. e. //
We
have seen
in Art.
favorable to
cases, //
where
is
its
and
Hence
if
the
number
A
,t*
The
"
_L_ _L
A2 z
A
_|_
-I-
two values of //
will
be
47
mining the relative precision of different series of observations is by comparing errors which occupy the same rela-
when
2a,
r,
therefore, n
is
number
the total
of errors, the
number
lying
between
limits
is
-\-
r and
also -.
Z
is
is
is
-,
fall
between
also
We
it
will
have, therefore,
V7r*S -r
from which
If
= r,
to find
we put // J
r.
/,
=p
corresponds to
then
2
1
Expanding the
= 0.47694
48
Now,
since
hr
it
= p = 0.47694 and
/ifj.
\/ 2
follows that
= 0.67457*
= - n roughly.
Hence
to
we compute
first
the
found.
will
It
the series
is
g~
Vn
Vx
or
T
i
e,,-(
'47694)
or
i
The
0.8
is
due
to take the
all
49
It is
where [J
is
may
expression for ^ in
be found as follows.
and J
-f-
dJ
\A
n
sum
the arithmetic
An
Greek letter rt
we have approximately
of the errors.
is
in the series is
A <f(J) dJ
*s
The sum
all
/+00
sum
of
A (p(A) dA
Hence
f+o
2/1
~7= /
VnJ
"
df
fe-'Vl
Indeed,
than
it is.
purpose
think
The
it
general custom
is,
however, to
T C
1
this
way
are as
=1.2533n
21
The
relations connecting
in the following
form
lar
= 0.6745 n
= 0.8453^
/JL,
r,
and
rj
are easily
remembered
51
If
we put
l
then, since //
= ri-f
is
-.
Now,
errors J assume
n(n-i)
j\
2h
21
Vx
4*4'
at a time,
is
(see Art. 6)
two
of n things,
is
all
is
number of combinations
that
we have
Squaring,
letting the
The number
(Art. 20),
"' P
of ft
LI
n*
is
//*.
Hence
finally
being the
is
-'
52
and
or
*=*('
= /jLii
when n
is
V2H'
very large.
average error
have
37
we
pro-
We
ceed
Let
Then
= /. 4.
n
'"
'
^^ w
in
TT
37.
Also, since
the error in
// is
a
~=
I/TJJ
Hence by
this
/7T
/
A/ -2 \f
K
iU
-I that
method of computing
mean limits
.A
is, //
7T
JJL
A/
f
the value of
//
is
>
53
i,
From
the equation
we may
equation
-*
r/
we must conclude
is
This
correct.
may
be
25.
"
says
The
the m.
s.
e.
or the
p. e. in
Gauss
largely a matter of taste.
so-called probable error, since it depends on
is
hypothesis,
banished
I,
for
my
part,
it
for
employed.
In this
m.
s.
e.
book the m.
and
p.
e. in
s.
e.
will
be used
in
54
We
is
Now,
if
dx
is
Fig.l
from
OX,
to the
falling
the
sum
or,
of
more
all
is
will
^ e~
h<t
Vn
limits of error.
Hence
known, by
for
a series
giving to
of observations
all
values from
-j-oo
whose quality is
oo and drawto
This curve
27.
is
_2~2
Since x
power, the
when x
Also,
= o,
55
-=o;
ax
that
is,
the
= o andj
x
axis of
is
Jy
-fax
an asymptote.
Again, since
-*
there
is
--
dx
t/jj.
a point of inflection
X
and the m.
flection.
s.
e. is
Also,
when
-:-=.
2
h^
fJ.
when x
= o, ^
dx*
is
in-
The
values of
//,
that
is,
diflfer-
\'
We
56
kj
^dx,
for values
Vj=.e"
7T
we have denoted by
of
extending.
oo
If, then, we
unity.
area
the
of
this
the
total
number of
curve
represent by
errors that occur in a series of observations, it follows from
from
-j-
oo
to
those limits.
Thus
if
is
to the right of
OA would
represent
of positive errors and
the area to the left of OA the numthe
number
AOX.
then the
number
of positive
OP
fyx dx
57
=
The
Law
\/~
y,
of Error applied
to
tions.
00.
\ Tt
tween
In
all
included within certain finite limits, and the probathe occurrence of an error beyond those limits is
of
bility
zero.
Practically, however, the extension of the limits of
error to
can make no appreciable difference in either
as
the
function
case,
<p(J) decreases so rapidly that we can
regard it as infinitesimal for large values of J in other
error
is
58
We
a, in
a series of observations,
2h
This
may
be put
form
in the
(/
and
is
The method
argument
that
is,
P
reason for arranging the table in this
convenient to compute
The
is
than
r
way
is
that
-.
r
it is
The
more
p.
I.
by deducting
Hence
tions in
in 10,000
any case
in taking the
We
are
duced by replacing
as found from
or 3//.
the change intro-
59
by
:=
as found from
law of error.
Hence
c=
Jl
2/1
-.(1+
\ - \
and taking a
$r,
=.
we
/*<*>
- A ** 3
tU
approximately.
I.,
//
,
"~^ 2 --9993)
= 1.001 Th
(
7T
Hence
We
made under
The
acters.
all
Now,
we
in
an ordinary series
we assume
vol.
ii.
60
We can certainly
its
is
more probable
If we could go back
instrument, observer, and conditions.
error
we
could
find
law in each case.
of
this
to the sources
Let us follow out this idea in a few simple cases and see to
what
leads
it
32.
where all errors are equally probwhere J can with the same probability asvalues between -f- a and
#, the extreme limits of
that
able
sume
all
is,
error.
Since a
is
the
maximum
error,
s* + a
dd=i,
I
/ -a
(p(A)
<p(A)
being constant,
and therefore
K^)=
20.
For the m.
s.
e. //
we have
= /r*
J -a
3
Also
/*
a.
2a
The
or
p. e. is
r+ r dA
/
J -r 2a
6l
and
that
is,
the p.
This
is
e. is
y=
constant
Hence by
the p.
Fig.3
find
e.,
The
be -.
we
definition
p. e.
would be repre-
o
p
x
sented in the figure by OP.
33. Next consider an error to arise from two independent
sources, x, y, each of which can with the same probability
assume all values between -f- a and
a, so that for the total
error J we have
number
oferroris
2a
2a
4<i
or
Z7 x Z( (ZO"
for y,
a,
a,
a
a
-f-
dJ,
dJ,
is
a
a
-(-
-f-\-
find the
magnitude
answer
-/
2dJ,
to
are favorall
possible
namely,
2dJ, ...
therefore
same way we
of possible causes
dJ,
-\-
dJ,
number
J
to be
of causes favorable
~-^-
62
Hence the
and J
that
-f- */J is
probability
<p(A)
dd
of
an error between
given by
is,
it
when J
lies
between
when J
lies
between o and
and o
4a
=- _
O//
For the m.
e.
s.
2a.
we have
r 3.
J
o
For the
-{-
r>
'
~+J
4
_ 2a
p. e.,
s*o
// o
// - r
A
-~
4
2a
-\-
J
y
2d!
= 2a
X
-f- x
Fig.4
P
34.
From
the preceding
we may
OP rep-
derive an important
practical point.
seventh place
never
is
in
error by
more than o
5.
Hence,
being the
maximum
error, the p.
63
of a log. as given in
e.
The interpolated
value at the greatest distance from a tabular value is the
mean between two tabular values. Its p. e., from Art. 33, is
the tables
is
0.25
in the
(2
Hence
the
seventh place.
1/2)
=0.15
of the log. of a
p. e.
sponding to
0.25
i,
Art. 7)
approx.
22 X 10
X mod.
Suppose now that we are computing
io
The
a chain of triangles
the base may
p. e. of
of its length.
Hence the error arising
1,000,000
from this source is 22 times that to be expected from the
be taken as
log. tables.
Again, the triangulation will be most exact, and
therefore the test most severe, when the angles of each triNow, the change in log sin 60
angle are equal to 60.
/r
of o".25-
p. e.
of log sin
which
Hence
60=
p. e. is 15
12.2
0.25
log. tables.
The common
practice
is
to
carry out to
and then drop the eighth place in stating the final result.
(See Struve, Arc du Mcridicn, vol. i. p. 94.)
35. If an error J arises from three independent sources
of the same kind, each of which can with the same probaa and -- a, then, the
bility assume all values between
64
maximum
to that
<p(A)
= -^
'
when J is between
-f-
when J is between
-f-
a and
a and
3#
'
lua
2
and
-J-
(2
^( J) = -3_
8rt
= ^=
Also
=2 A
Jo
and
/;
The curve
the figure
Sl>*
OA = OB
OC=ODa
30
PC
36.
\
,
A consideration
and B. The
sented by OP.
at
p. e. is repre-
show that the more numerous the sources of error assumed the nearer we approach the results obtained from
the Gaussian law of error. Thus for
will
one source
two sources
three sources
Gaussian law
T
-
1= 0.87
T
-
= 0.88
= 0.87
0.72
= 0.71
= 0.67
i.oo
=10.85
The forms
so
we
due to Zachariae.
37.
Law
of Error.
The same
made on each
observations were
results for the first
and
last
of twenty-four days.
0*30
0?10
0?35
Fig.7
0?20
* Coast
0*30
0?25
Surrey Report,
1870,
appendix
21.
The
In the
66
38.
General Conclusions.
On
-=
e~^^
V 7T
When
in a series of
observations
we
we can
we have exhausted
is
this
There was
in-
it
67
of the results was made, and the law of daily change discovered, which gave a means of applying a further correc-
tion.
new
correction,
later,
after taking
account of
this
of Observations.
39. For purposes of reduction observations may be divided into two classes those which are independent, being
subject to no conditions except those fixed by the observations themselves, and those which are subject to certain
conditions outside of the observations, as well as to the conIn the former class, beditions fixed by the observations.
fore
set of
values
is
The angle
BOC could
AOC=AO + BOC
The unknown
BOC
But
if
A OB, A OC,
andno
10
68
knowns.
(3) Condition observations.
CHAPTER
III.
UNKNOWN QUANTITY.
IN the application of the ideal formulas of Chapter II. to
an actual series of observations we shall begin with a single
quantity which has been directly observed. We shall con-
two
sider
cases
when
first,
all
when they
are not
all
of equal
quality.
40.
Mean.
We
values
value
Fof
lt
arithmetic
It
the Arithmetic
M^
mean of
by making
minimum.
shown
that the
is,
same
the
(2)
and
V is
to be found from
,'
that
is,
+ .+
+ *>'=: a min.
(3)
from
.
+(F-J/ = amin.
M )'
(4)
70
differentiation of (4)
By
and
(5)
(6)
41.
As
M=o
values M
the observed
(7)
are
often
numerically
work
of
differences
/t
/,...
respectively.
Then
M X =l
t
M,-X' = l
(i)
jf.--r=4
By
addition,
n
X'-{-.v' suppose.
Hence
all
that
we have
small quantities
to the result.
/2
to
.
(2)
/,
DIRECT OBSERVATIONS.
71
find the
It
80
177"
21'
5".
177
21'
7".35
177
2l'
4".
28
mean.
sufficient to find the
is
mean
of the seconds
and carry
in the
degrees
In
least
may
be months, to complete
it.
In long computations
it
is
better for
even
if
control
is
essential.
A control
M=o
sum of the positive residuals should be
of the negative residuals.
If, however, in finding the arithmetic mean, the sum [J/]
of the observed quantities was not exactly divisible by their
that
is,
that the
equal to the
number
sum
the
mean
differed
from
72
DIRECT OBSERVATIONS.
73
The degree
unknown
is
shown by
its
T of
44
V and the
vn instead of the true values
4,
4,.
i\,
?'
Now,
y-v =M = r- 4
V-v = M^= 7-- 4
1
(i)
By
addition,
remembering
that [?']=: o,
nV=nT-[J]
Substitute for
Fin equations
nv
= (n
(2)
and
4
i)4
-4 -)-(_ 1)4-
WT'.,=:
Squaring,
wX'=(-i)'4'
v,- =
(i)
+ - 4' +
4' +
0*4' +
(
.-2(-i)44-.
- 2( - 04 J -
By
probable,
= (and
//'
_^
= _t^
:
w
s.
e.
(3)
i
of an observation.
74
Now, from
Art.
19,
/'
s. e.
of the arithmetic
mean
of n observa-
The
-^ -
result, //
For the
//
(4)
would be a close approximation to the m. s. e. of an observaIt is, however, not satisfactory, from the fact that it
i, which it does
ought to become indeterminate when n
not.
For when
i, z>
o, and unless the denominator of
(4) is equal to o, // would be equal to o; that is, the first
observation would give the true value of the unknown,
which is absurd. Hence we should expect the formula to
be of the form
tion.
,,_.
n-l
//)
/
'
equal to //
4/^_
n
r
We
s.
e.
have, therefore,
ofu
=n
is
DIRECT OBSERVATIONS.
and when n
is
75
very large
/
/ [w]
z:
'-V^'V -fi&=$
that
is,
the
mean uncertainty
of /*
T)
\'2(n
45.
From
is
existing
between the
where
/>
46. If
of
4/2
= 0.6745
we consider
the m.
nearly.
s. e.
<j.
of the arithmetic
;/
fa-\'n
lows
and u*=n
!-
have
mean
We
By
1/2
mean
mean
V, the results
neatly as
fol-
?6
Error of
F=I (4 + 4 +
+ 4)
and
,_
Vn
Again,
A=TM
V+JL-.M
'
Vn
and
[v~\
o,
Formula.
From
first
the equation
1034.
DIRECT OBSERVATIONS.
77
to sign,
;/,
1.2533
=
For a demonstration of
\y
nearly.
this
in
form
of
observations
we have
mean
respectively
0-8453
Vn(n
r
The mode
ceding
is
--
0.8453
-
\y
of deriving Peters' formula given in the preapproximate, and the formula itself is not very
/8
n.
Thus when
= 2, if ^/denotes the
d
^, and by Peters' formula
it
V2
fourth greater.
the same ratio.
d
1.25
-=.,
which
is
one-
V2
errors are in
_
~
r
T
A/2 n
As
9SS
0.8548
\v_
results
To
= 0.6745
= 0.6745
yl.
n
= 0.8453 -7=
vn(n
r"
_ l}
Q
1
=a84
53,,
/->
v;/
i)
have computed
tables containing the values of the coefficients of V[^'] and
[v in these equations for values of n from 2 to 100.
(See
Appendix, Tables II., III.)
If Bessel's formula is used compute first [vv~], then V[^ ]
can be taken from a table of squares closely enough. This
square-root number multiplied by the number in Table II..
corresponding to the given value of n gives the p. e. sought.
If Peters' formula is used multiply the sum of the residuals,
without regard to sign, by the numbers in Table III. corI
n.
DIRECT OBSERVATIONS.
79
Control of
49.
derivation of
[i'?>\
We
have
vn =
V-M
= n F - 2 F[J/ + [>/']
2
[w]
values of J/
(i)
nV=Jf
since
The
= [j/']_[j/]r.
= [//]-[/>'
50.
A connection between the p. e. of a single observaand the greatest error committed in the series may be
established approximately by the aid of the principle proved
in Art. 30.
There we saw that in a large series the actual
errors may be expected to range between zero and 4 or
5 times the p. e. of an observation.
If, then, we find from
cision.
tion
The
probability of
its
may
assert
more than
is
only
about T oVo.
The same
the p.
e.
in
80
measured
results will
their difference
may
maximum
error.
Hence,
<p(d) in a series,
P
where
P, Q, R,
QJ? -f
RJ -
we may
.
Expanding
write
are constants.
a denotes the maximum error, then, since the probaof the occurrence of an error between the limits
a is certainty, we have, taking two terms of the
-\-a and
If
bility
series,
f
j
-<
Also, since a
From
of
is
the
maximum
P and Q may
be found
is
zero.
in
terms
a.
To
find the p. e. r
we have by
Pr
or
- Qr>
Substituting for
P and Q
their values,
and solving
for
r,
we
find
- nearly
,
that
is,
the p.
e.
is
approximately \ of the
maximum
error, or
DIRECT OBSERVATIONS.
closer approximation
terms of the series for ^(J).
81
We
= - nearly.
4
See note by Capt. Basevi, R.E., in G. T. Survey of India,
also Helmert in Zeitschr.fiir Verviess., vol. vi.
vol. iv.
;
Ex,
We
now apply
shall
42 to find the m.
observation.
s. e.
and
p.
(i)
(a)
From
the
sum
= |/_If^L
i)
n(n
'
21
ra
or from Table
once
II. at
2O
= 0.026
= 0.6745 X O.O26
= 0.017
;-
=0.525
0.033
= 0.017
From the sum \v of the residuals (Art. 47):
The multiplier in Table III. corresponding to the number
(b)
(2)
The
p.
e.
of a
From Tables
II.
ra
0.06
order of magnitude.
They
0.041
0.09
0.009.
III. directly:
r=2.O2
0.22
is
single observation.
and
0.23
21
= 2.02 X 0.009
= o.oiS
0.19
0.06
=0.082
in
0.16
0.16
o.n
o.n
o.io
o.io
o.io
0.05
0.04
0.04
0.04
0.04
0.02
o.oi
The
Check
(ft).
See Art.
50.
Range
the different
well.
are:
0.09
e.
r of a
82
51.
now
We
test
- residuals is
9.
(2)
The sum
- residuals
(3)
is
of the
The sum
and of the
-f-
residuals
is
1.03,
0.99.
residuals
is
-f-
residuals
is
1417,
1339.
To find the
(4) The p. e. of a single observation is 0.08.
number of observations we should expect whose residual
errors are not greater than
'
argument
0.08
1.25
and
o. 10,
we
find 0.60.
enter Table
I.
with the
This multiplied by 21
From
The preceding
table
results
Limits of Error.
are
collected
in
the following
DIRECT OBSERVATIONS.
Table I., it will be remembered, is founded on the supposition that the number of observations in a given set is
very large. In our example the number is only 21. Peraccordance between the number of errors given by
theory and the number given by observation is, therefore,
not to be expected.
For longer examples of this kind see
Chauvenet's Least Squares, p. 489 Airy's Theory of Errors
of Observations, third ed., appendix.
Comparisons between the number of errors within given
limits that actually occur in a series of observations and the
number to be expected from theory in the same series show
the degree of confidence we may place in the law of error.
It is the final criterion, and forms the second part of the
proof of the law as stated in Art. 11.
The law of error has been so thoroughly tested in this
fect
way, so
the sciences of observation (for which, inwas framed) are concerned, that if in a series of
far as
deed, it
observations
Limits.
84
The agreement,
Prof.
theoretical number.
For another
illustration see
found to apply
reasonably well to other phenomena, such, for example, as
statistical questions, guesses, etc., it has been often rashly
assumed to be of universal application and when prediction
and experience are found not to agree, the validity of the
law in any case has been as rashly impugned.
In the fundamental investigation in Art. 15 the hypotheses there made are satisfied by other functions of the
measured values besides the arithmetic mean. Thus, tor
example, taking the geometric mean g, the resulting law of
is
where
DIRECT OBSERVATIONS.
ever, no
85
in
its
own
territory.
misapplication
where
it
would appear
is
it ought to
apply
"
on
essay
Target-ShootF. \V. Herschel's Familiar Lectures on Scientific
to be found
directly,
"
in Sir J.
ing
in
the
Subjects.
53.
of Precision.
We
series
With
ideal
systematic
and
in
It is
common
mistake to over-
s. e.
and
p. e. as
For example,
are made.
small.
The m.
s. e.
86
The conclusion
arrived
The
difficulty
may
derivation of the m.
ties
lf
M^
s. e.
we
Also
//
= n\vv\
we suppose each of
changed by the same amount
Now,
if
= V-M
the residual errors will be the
*
See, for
same
example, G. T. Survey
as before.
of India^
vol.
ii.
DIRECT OBSERVATIONS.
Hence
m.
s. e.
if
is
87
unchanged, and we
makes no allowance
is
result as small as
fore,
seem
to be in the
we
way
please.
Nothing would, thereof our getting an exact result,
and that we could do as good work with a rude or imperfect instrument as with a good one by sufficiently increasing
the
number
of observations.
series of
nearer
is
circle Prof.
Rogers
p. e.
.O26.
measure
88
known
The explanation
is,
as inti-
if
we have
be mentioned.
the p.
e. is
of levels between
of a lake
we
Amcr. Acad,
vii.
DIRECT OBSERVATIONS.
89
take the value resulting from the observations, unless, indeed, it came out with a negative sign, and then its unreliable character would be evident.
Constant Error.
54.
in the
two
error
us to notice an example or
preceding
of the detection and treatment of this great bugbear of
articles lead
observation.
We
jump
at conclusions.
olite,
the
if
embedded
instrument
in
at
night, using
Emle Lake
At U.
S.
Lake Survey
station
uniform sunshine and clear atmosphere this twist seemed to be quite regular, and at the rate of
about ant second of arc per minute of tite, reaching a maximum about 7 P.M. and a minimum
about 7 A.M., during the month of July. On partially cloudy days there was no regularity in the
twist, being sometimes in one direction and again in the opposite.''
9O
work
the
mean
mode
This
of procedure
is
in
See Art.
The
2.
two
effort to
DIRECT OBSERVATIONS.
91
first
The
55. Necessary Closeness of Computation.
number of observations necessary to the proper determination of a quantity
the
m.
s.
method
e.
may
be approximated to by referring to
If, after planning the
of these observations.
stant error," we find that increasing the number of observations beyond a certain limit does not sensibly affect the
m.s. e. we ma}' conclude that we have a sufficient number.
,
of F, then J
-\- c,
J^-\-c,
Jn be the errors
Let J p J3
Jn -\-c are the errors of V-\-c.
,
Hence
and
/V+c=#p|i-|
V
Now, we may
13
between
// r
and
92
Hv to be
Hence
--^.
troduced by computing
in
--
<
"
2 fiy
that
m.
is,
when
100
is,
roughly,
of the
s. e.
Mean.
shown
in Art. 17 that
if the
directly
of a quantity are of different quality, the most probable value is found by multiplying each residual error of observation by the reciprocal of
its m. s. e., and making the sum of the squares of the products a minimum that is, with the usual notation,
It
M^
observed values
Mv
/V
+ /V!+
..
+^i = amin.
(i)
!*n
or
=a
By
differentiation
min
and reduction,
(3)
We
it.
Let
p^p
DIRECT OBSERVATIONS.
be the numerical parts of/v /v
of the type
is
/v
(unit of measure)
93
I?
then equation
(3)
Also, since
may be
f*i
written
P*
t**
same proportion
to
/V /V
that
is,
f'-n
if/>,'A
(5)
where
is an arbitrary constant.
Let us look at this question from another point of
It is in accordance with our fundamental assumpview.
tions that observations of a quantity made under the same
conditions, so that there is no a priori reason for choosing
one before another, are of the same quality. They require
//
57.
values being
of
the
same
quality, the
94
Fof
is,
unknown
the
is
by
y _ sum
of the values of
~sum
of
number
A+A+
which
of the
is
same form
The numbers
the sets
all
+A
as Eq. 4, above.
are called the weights, or,
A ..... pn
may be
Art. 56
one
lent
If the
that
By
is,
=a
is
found from
min.
from
differentiation,
p (V
l
whence
If
value by
the
its
weight,
sum of the
and
weights.
V leads
to the rule
is
DIRECT OBSERVATIONS.
As
arithmetic
mean
95
it is
(Art. 40),
evi-
dently simpler
weighted mean directly
this
rule
rather
than
from
the
minimum
by
equation.
If the observed values J/are numerically large we may
lighten the arithmetical work by finding V by the method
of Art. 41. Proceeding as there indicated, we have
in practice to find the
= X'
58.
-\-
x" suppose
Standard.
VA
Now,
the m.
since
s. e.
/Jt t
of
VA
A/A
tt
are the m.
M^p^M^/p^
same quantity
Hence if a
weights A A>
/*
s. e.
of
M Mv
n,
n.
serits
of observed values
A>
t ,
M^
J/,
=v
V-M, =
weight
V-M =v
n
"
"
i>,
V.
have
tJie
same standard by
VA respectively.
to find the
lt
g6
to the
Reducing
equations
A/A
v- VAM, = A/A^
Vpn v
A/A,
^4=
is,
+ (A/A fO* +
+ VA *)' = a
rain.
by making
V- A/A ^)
this equation, \ve find, as before,
Reducing
59.
If
A> A
s. e.
known.
These relations suggest that
define
of the
//
as the m.
s. e.
it
would be convenient
to
weight unity.
We
whose m. s. e. is /x
60. Control of the Weighted Mean.
may be
written
[pv]
=o
Eq.
2,
Art. 57,
DIRECT OBSERVATIONS.
the
weight
is
97
residual
correctly,
error by
its
equal to zero.
i.
Deduce
the
relation
[/^]
directly.
Ex.
2.
its
Find the most probable value of the velocity of light from the
298000
298500
1000
kil.
"
looo
"
1000
"
100
299930
{Amer. Jour.
the
"
"
"
"
200
299990
"
300100
kil.
(Art. 59.)
Direct solution.
according
to
Art. 57.
Assume X'
298000
DIRECT OBSERVATIONS.
61.
99
m.
s. e.
where n
Y of
is
[/>]
Since
observa-
Hence
the
the m.
s. e.
(standard observation).
to Art. 58, the value of }t may be found by
for v lt i\, ... in the formulas
?' 2
.
.
7-,,
According
writing \X/>,
VA
derived for observations of the same weight. Hence, substituting in Bessel's and in Peters' formulas Arts. 43, 47, we
>
have
:=
n
1.2533
\/n(n
i)
and therefore
..3533-
= up.
is
vt
Hence
VM,
=V-M
weight/,
"
100
By
addition,
\_pirj]
= [>] V* -
since
In
is
\_pMM\
squares
is
best.
found that
on
for the
small.
F. of the G. T.
o". 050027, o.O4997i, o'. 050019, o'. 050079, o1'". 050021, o. 050011. The numbers of measures in these determinations were 6, 6, 15, 15, 8, 8 respectively.
Taking the numbers of measures as the weights of the respective determination-, required the most probable value of the space and its p. e.
The direct solution presents no difficulty. The value of
may be found
We
The m.
s. e.
or p.
e.
follows
61.
v.
in
this case
DIRECT OBSERVATIONS.
Assume X'
/
= 0.049971
101
102
from the original observations rather than from any combinations of them.
The weighted mean value V would evidently be the
same whether computed from the partial means or from the
original observations.
63.
ples of the
a X, a^X,
M^
lt
'
same unknown
;
a H X, where a^ a v
.
that
.
is,
The values
of
?,
an
#*
weight.
MM
of
is^,
is
the m.
since
then,
.,
s.
e.
of an observation, that
the
m.
s. e.
of
_'
a
is
is,
^,of
a
a^
fi.
a,
Hence taking
the weighted
mean
is
the weight of X,
*_=
P*
['J
Cotes,* in solving this problem, reasons that, since for
the greater a is, the less is the error of
same error of
X, we may take the coefficients a to express the relative
weights of the values of X. Now, placing the coefficients
as weights along a straight line at distances
a a^
the
t ,
_', ^?,
line,
a,
a,
*
Harmonia Mensurarum.
Stockholm,
1860.)
Cambridge,
1722.
DIRECT OBSERVATION'S.
103
and therefore
Ex. To
test the
power
ft.)
of the
Ben More, Scotland. It was so arranged that when projected against the
sky a fine vertical line of light, the breadth of which was regulated by the
sliding of a board, was shown to the observer. The breadth of this opening
was varied by half-inches from i^ in. to 6 in. during the observations; which
were as follows *
:
No. of
obs.
104
Let
i
inch.
9.50
= vi
5.5^8.50 =
5
=
6.50 =
6.66 =
'
From
the
preceding
or
making
the
sum
find
2/4
2.
2/6
X=
1.58
2
weight 6
JT=
1.55
weight
5-5
2/7
z/8
2/8
z/ 10
--
+
1.58X6'
-
1-55
1-55-
X5-5 2 +
- 9.50)*
(6Awe
2.5
weighted mean
4.17
X
X
1.5 X
2/3
2/8
=
3.33 =
3.17 =
1.66 =
3 A"
we have
weights
.'.
3.5^5.84 =
2/3
8.00
4.5 Af
(5. 5 A"
8.
so)
minimum,
that
is,
a min.
by differentiation that
AT=i.55
as before.
The practical rule following from either method is the same, and may be
stated thus
Multiply each observation equation by the coefficient of A" in
:
that equation,
of X.
We
The
In other words,
it
minimum
was
that
we cannot
squares
we
is
X= 1.58
and take the mean of these values as the value of x, because these equations
should be written strictly
in
the
the
DIRECT OBSERVATIONS.
IO5
Precision
Quantities.
/j.
F=a M + a M +
#
av
a n being constants.
This has been already solved
.
(i)
shown
+an Af
that
2
= [a
r
2
[t.
Art.
in
19,
where
was
it
an
]
On
J
and
F;
+.(-*/. +40
T= F-\- J,
+a J
+.
then, since
=^ J +
1
+ 4)+-
^.2
we have
.*.
ff
= aW + af4+. .+2^,44+.
lt
be
In
forming
number
as many
all
values of
possible
J,-/,,
JJ
2
.,
the
of values
-j-
as
assume
[J JS ]
1
= [J J ]=.
1
=o
Hence
^=[>y]
(3)
Io6
Ex. i. The Kevveenaw Base was measured with two measuring tubes
placed end to end in succession. Tube i was placed in position 967 times,
and tube 2 966 times. Given the p. e. of the length of tube i
o*.ooo34,
and of tube 2
o'".ooo37, find the p. e. in the length of the line arising
from the uncertainties in the length of the tubes.
from tube i
from tube 2
[ p. e.
p. e.
.'.
p. e. of line
to be
o'.o3
Show
that the p.
e.
[ p. e.
of
measurement
of
tube
at
o*".i36.
'-^
1/94
p. e.
0.03
-
0.136
Attention is called to these two problems, from the importance of the principles illustrated. In Ex. i the p. e. of
a tube was multiplied by the whole number of tubes to find
the p. e. of the base from that cause, for the reason that
with whatever error the tube is affected it is cumulative
3.
The m.
quantities Mi,
each being
Mi
ju, is
If
of
aM\,
s. e.
Vaju.
being the m.
is,
sum
of
Mi
s. e.
of
be expected to
M\ and
a a constant,
is
Explain.
form,
/<
of the
a ; that
the function
65.
in the linear
7,
s. e.
but the m.
to aju,
may
3,
we
F
first
Art. 64.
whose m.
reduce
it
Thus,
s. e.
is
required
is
not
if
DIRECT OBSERVATIONS.
M^-^-dM^,
for
of
J/p
result
will
J/,,
if
we
write
Then
first
T= f +
or
Error of
where
This expression
HP
The
still
is
= a?ti? -f
tf,V/,*
-f
(i),
Art. 64.
a^i'-n
fay ]
of the m.
s. e.
Hence
of a function of
quantities which are themselves functions of the same observed quantities may be readily reduced to the form of
The whole point is to express the error of F
Eq. i.
as a linear function of the errors of the independently ob-
served quantities
J/,, J/,
J\In
from differentiating the function equation directly, as has been already pointed out in Art. /.
It
is
Ex.
i.
angle
useful
If /i,,
ABCD,
it.,
to
note
are the m.
find the
m.
s. e.
that
s. e.
Eq.
of the
[Here
F-A^M-,
.'.
by differentiation
,/F=Jlf
<Uf.1
J/,,/J/,
and
HP15
results
/I/, -//..-
Af-i'n
'-\
BC of
a rect-
108
Ex. 2. The expansions of the steel and zinc bars of tube i of the Repsold
base apparatus of the U. S. Lake Survey for i Fahr. are approximately
Show
5=
0.0248
o.oooi
0.0617
0.0003
'
that
S --2
nearly.
400
F=
[For
and
(p. e.)-
(0.0001)-
2
4
(o.ooos)
Ex. 3/iThe base b and the adjacent angles A, C of a triangle ABC are
measured. If their m. s. e. are respectively //*, J.IA, Uc, find the m. s. e. of the
angle B and of the side a.
To
find
We
/.IB-
have
y?
where
xSo
Hence,
To
MB"- HA* +
find
jit
a.
= b- Asin B
,
By
l-ic'-
sin
differentiation,
sin
da =.
sin
--db+l>
(C
sin
sin*
e)
-
sin
dA
a cot
sin
dC
and therefore
"a"
Ex.
4.
Mb, HA,
We
sin 2
-^"T-f>t">
sin^
!>
sin 2
(C
s)
I-
have
as follows
MB
sin 2 i"
-T^B
sin
in the
s. e.
f.i
a cot
A,B
sin
/'C
of a triangle with m.
of the side a.
sin
B sm
s.
(i)
Take logarithms
of both
log a
members.
= log
e.
+ log
Then
sin
log sin
(2)
DIRECT OBSERVATIONS.
(a)
By
109
differentiation.
da
db
a cot
sin
"dA
u cot
B sin
"dB
Hence
a1
Ha"
Hb*
o'
cof2 ,-/
sin^i"//,!
<x
(3)
0"
If,
as
is
MA
= MB = H,
and
/m,
=o
then
f.t
= a sin
i"
(b)
tiating (2)
= db db +
S a da
+ cotfB
V'cot'M
f.i
in
Chapter
A dA
f)
t)
(4)
we have by
I.,
differen-
dB
(5)
where ft a di> are the differences corresponding to one unit for the numbers a
and b in a table of logarithms, and dA, SB are the differences for i" for the
Hence
angles A and B in a table of log sines.
,
'
and
(3)
may be used
(6)
5.
is
to
Z A = 57
Z B 67
The
Middle-Traverse
side
Taking
16894.9 yards.
We
fti,
Then
HA
= o".30
15' 39". 2
/IB
o".29
Id. as
is
have
a
04' 51". 4
log
(a
(see Ex.
da)
i,
log
Art.
7),
(b
dt>)
the
sin
sin
log sin (A
+ dA)
log sin
in
(B + dB)
units
of the
(b
colog sin
.
log a
= 4.2277556 +
= 9.9239892 +
+ dA)
(B + dB)
0.0351398
db)
log sin (A
+ $ a da
4.
4.1868846,
and 283 da
Since log a
257 db
14
dA
dB
1868846 + 257 db + 14 dA
257 db + 14 dA
^dB.
table.
and 283
is
dB
IIO
Hence
,.
and
(.05)'
(?g)'
Ha
0.0024
0.05
=
=
o".42
(.30)
(--)'
Also
V'(o.29)
(0.30)-
Miscellaneous Examples.
>"
66.
Ex.
The
2.
p. e.
.~^-^
triangle-error
is
;-
of an angle of a triangle
\ 3, all of
[Error
Ex.
it is
The expansion
3.
$a
is
r;
show
that the p.
e.
of the
Fahr.
= 180 - (A
of a bar for
+ Q.]
C.
is
ga
9/7
show that
for
$r.
Ex.
The length
4.
was a
measurement
After
>'i.
it
was
;- 2
Show
that the
th
measure, the length being supposed to change uniformly
length of the
witli the distance measured, is
a+
-(i
]da
a
is
and hence
Ex.
ference
db.
common
p. e.
/>,
e. is
e.
{/ r^
of the
p. e. of
(b
-\
a) is
follows.
5.
is
is
x)
It
,'^
da
if
[For
6-
-\
-(r^
mean
form da +
in the
(db
da),
'
r2 2 ).
of two observations
each observation
is
0.477 d.
whose
dif-
DIRECT OBSERVATIONS.
Ill
Ex. (i. The line Monadnock-Gunstock (94469 m.) was computed from the
Massachusetts Base (17326 w.) through the intervening triangulation. The
om .3i7, and the p. e. of
p. e. of the line arising from the triangulation is
the base
is
0^.0358
[p.e.
Ex.
7.
0.0358)*
1325
ft.
15
that the p.
e.
15
ft.
bar at 32
bar at 32
of the base
o.372]
to sea-level is
n".3i4
4-
+(0.317)-
and
show
f/(^
o'.42i
o'.oooi2
179'". 95438
o.45o.
is
is
it
We
o'.45o.
but whichever
multiply
it
is,
it
is
V-+2
H being
the m.
s. e.
9.
HI
n-z
of a single observation.
[Latitude
Ex.
'
90
f (?
?')!
results,
h.
m.
s.
Omaha
Springfield east of
show
s.
0.23
0.06
25 08.69
o.
58 37.34
0.26
.11-
= 0.26]
that
E.\. 10.
.23-
.06-
of
moon =
305879
moon
and
mass
prove
mass of earth
of
mass
2299
309635
[For (305879
22-1)
X ^
c
of earth
44
309635
2299]
2271
112
Ex.
ii.
= p.
= p.
of a pointing at a signal,
of a reading of the limb of the instrument,
=. error of graduation of the arc read on,
r^
r-i
then,
show
assuming
if
e.
e.
that these result from the only sources of error not eliminated,
times and n readings taken in each
position, that
p. e. of
angle
--
+ rj)
/2(?V
c2
Ex.
12.
The distance
[For distance o
=^
..(p.e.)
Ex.
p. e.
a graduated line-measure
of the
mean
of two results
mm
is
readings].
is
read with
equal to the
+ second
(first
mm on
//
rdg.) at o
+ second
(first
rdg.) at
mm
\
=i{4(p.e.) ofardg.}]
13.
In the comparison of a
side by side and read with a micrometer, the p. e. of a single micrometer reading being a, show that the p. e. of the difference of the results of M combined
[For
and
a reading
p. e.
of a
p. e.
of
of
two measurements)
/2
4/ -a.
=a
combined measurement
mean
is
=a
combined measurements
of n
etc.]
Vn
Ex.
same
14.
theodolite
precision.
[p. e. of
rj
= error of reading
(p. e.
error of pointing
of ptg.) 5 etc.]
,
Vn
DIRECT OBSERVATIONS.
Ex,
the m.
show
15.
s. e.
values of
b* is b t
V[>
].
[Discrepancy
db*-l>
and
Ex.
and
/a
t\
show
p. e.
,,-)
A-(/<,*,)
= n",
= /<-]
was af
rj, and
chronometer
the
of the correction to
Correction
=a
.*.
Ex.
chronometer
at
an
.v
"JT.
JT
>
of
A) at time
(/
n
ra
or.
+ /
-r^
.//I'-vrI/
M"
2 ->./-
Similarly p.
e.
of
n:
l-i~
iS.
to
when
.-.
/-,
^i
x cos a = A
x sin a = /2
and
p.e. of
I*
Given
17.
find p. e. of
from
p. e.
ai
-\
If
A)
/'.
I",-.
OD
/i
interpolated time
.'.
(;<
ra y
the
find
s
/'i
/'./W+/<3 +
/ 2 it
= / suppose
d\Xdl> =dl
2
At the time
was a-f
16.
time
at the
/>]A
fist**
or
is
113
=r =
2
i**,
for
=i
i TV'
example, then r
+ OB)
+ TV
o*.S,
when
//
= one
micron.
It
quantity
ever,
OD
114
Ex.
Given the
19.
p. e. of
.r
be r; find the
to
r aj
log x
L
log x
p. e.
'
mod.
~l
a-
Ex.
of log x.
p. e.
mod.
_ dx
x
measurement of
was
rections
Show
o"*.o332.
that the
[p. e.
(2165
p. e. of the
is o'*.o358.
o^.oooooss)
-\-
V 2165)" -\-(om
(o".oooi27
0332)-
o>.035S]
Ex. 21. Given the length of the Massachusetts base to be 17326. 3763
o m .O3$&; show that the corresponding value of the p. e. of its logarithm is
8.973 in units of the seventh place of decimals.
log
0.0358)
(b
log b
log mod.
See Art.
(0.0358)
7.
9.6377843
log b
4.2387077
0.0000008973 3.9529596
Ex.
22.
The m.
decimal place
is
s. e.
10.6
of the log. of a
(N +
/'-log
and
Ex.
m.
e.
s.
= log N + ~rr
)=
(N +
TV-
/'
/mod.Y
)
10.6
T
//
number.
v)
mod.
.
67.
number
log
-5-
io 7
4iooooJ
Examplc of Weighting.
i.
DAE are
The weights
3, 3,
respectively
-
wt.
ti
ill
+ -,
=-+3
3
i
.'.
wt.
-0.6
-J
CAD,
DIRECT OBSERVATIONS.
Ex.
-u
,
r~
.J!'
X = a xi +
<ix.2
P the
x n and
anxn
and
p\,
p-i,
pn
^ ol
\ ^"'^.^
>
r-!*
'
O.
If
2.
of xi, xi,
115
-p
f/'
3095313
Mars
Show
the
3078456
that,
of
common
the
in
unit.
10104
Mass
of
Mars
3093500
Ex.
3435
On
4.
3295
m -is measured
anci
found
to
be
far as the
To
find P.
is
concerned.
m
(i
2'"
= (o
2"'
(o
i'"
-i
1113
- - - =
P=
P = + 2 = 2 and
3
=
the weight of
f, show that
=
*
of
weight
log
i
E.\.
5.
Ex.
6.
Given
If
.v
.\
= y and
weight of
Ex.
7.
Given the
.V
c-
Washington and
Key West,
m.
1873, Dec. 24,
Dec.
Dec.
Dec.
show
s.
s.
19 01.42
0.044
26,
1.37
.037
30,
1.38
.036
31,
1.45
.036
60
.046
1874, Jan.
9,
Jan.
10,
1.55
.045
Jan.
ii,
19 01.57
0.047
i.
that
weighted mean
weighted mean of
weighted mean of
last
=
first
m.
s.
19 01.460
first.
19 01.573
t.
0.016
0.019
0.027
Il6
Ex.
Craney
8.
Kent
Base
Craney Id. Base
Show
Kent
Id.
=
=
w.
Id.
the
26758.432
0.38
26758.176
0.43
that
(1)
(2)
of junction line
=
=
m.
m.
0.256
26758.32
.V
A,
'
0.57
0.028
Ex. g. In latitude work with the zenith telescope, if n north stars are
combined with s south stars, giving ns pairs, to find the weight of the combination, that of an ordinary pair, one north and one south, being unity.
[Let // = the m. s. e. of an observation of one north star or of one south
star.
mean
of
south
2ns
.'.
"The
n +
~[
.
weight of
fi>
In one
3X4
27only.
In this and
al!
similar cases
treated the
whole com-
3~r4
bination as one pair that is, I inserted in the blank provided the half-sum of
mean of the declinations of north stars ;md of the mean of the declinations
;
the
of south stars,
method."
For a
This
is
Ex.
north
10.
star,
If
and immediately
resulting latitude
is
is
the angle
Walworth-Palmyra-
74 25' O4".6ii
o".22, mean of 24 results
required the most probable value of the angle and its probable error.
;
DIRECT OBSERVATIONS.
[With first theodolite p. e. of a single obs.
Wiih second theodolite p. e. of a single obs.
117
= o".2()
= o".22
=
=
16
^'24
i".i6
i".oS
Let a single result with the first theodolite be taken as unit of weight,
then mean of 16 results has weight 16.
Let a single result with the second theodolite have a weight/, re f erred to
unit as the first, then mean of 24 results has weight 24 p.
The value
of/ is found from the relation
the
same
Vi.oS/
Also
and
weight of this value
r".429
16
+ 24 /
16
o".6n
4-
16
+ 24/
Note.
If, instead of being two measurements of the same angle, the above
were the measurements of two angles side by side, then
total
angle
148
50' io".O4o
because, no matier how much better one is measured than the other, we can
do nothing but take the sum of the two values. The weight P of the result
would be found from
.v. 12. An
angle is measured n times with a repeating theodolite, and
also n times with a non-repeating theodolite, the precision of a single reading
and of a single pointing being the same in both cases; compare the weights of
the results.
[ //i, /'a
the m.
s. e.
(pointing
.'.
and (m.
s.
(m.
e.)-
s.
of
e.)-
of
mean
reading)
(pointing
one measurement
of
;;
measurements
4-
= 2//,
=
reading)
5
(2//r
2// 2
2// 2
2
).
(pointing
.'.
(m.
s.
e.)-
of
and (m.
s.
reading)
pointing
pointing
pointing
pointing
(pointing 4- reading)
times the angle = 2/<, 2 + 2ftie.)-
of the angle
(aw//,
4-
2// 7
Il8
and hence
the
method
would seem
it
of reiteration.
that the
method of
This advantage
is
repetition is to be preferred to
so
much
less, the
smaller
is
-^
is, the more the precision of the circle reading increases in proportion to
the precision of the pointing.
This result is contradicted by experience so much so that in all of the
leading surveys repeating theodolites are no longer used in primary work.
that
Where, then, is the fault? Is the theory of least squares false? By no means.
have only another example of a point which occurs over and over again,
and which is so apt to be overlooked. (See Arts. 53, 54.)
The result obtained is true on the hypothesis that only accidental errors
enter.
We have assumed a perfect instrument But the instrument-maker
cannot give what the geometer demands. From various mechanical reasons
We
been devised
of getting rid of
it is
it,
Ex.
first
13.
An
made
are
angle
is
n, repetitions,
made n^ repetitions, // 2
Show that if
2 are
i,
m.
s. e.
of
NOTE
I.
far as
When
WEIGHTING OF OBSERVATIONS.
19
number
same
result.
approximately
is
only
true.
a millimeter space,
looo.
0.40,
1000.3
-33>
mean
mean
of 20 readings.
of 30 readings.
To find the weighted mean of these two sets of measurements we may proceed in two ways. The number' of results in the first measurement is 20, and the number in the
second is 30. Hence, taking the weights proportional to
the
number
of results, the
mean
20
p. e.
of the
+ 30
to
40'
= IOQ0.220
that
is,
as 1080 to 1600.
0.33,
and
33'
wide ranges, giving a small weight. A great deal, theredepends on the judgment of the computer in deciding
what weight is to be given, it being constantly kept in mind
that the strict formulas which are correct in an ideal case
must not be pressed too far in practice. Thus in the second
set of observations above the first three results were 999.8,
fore,
120
The
but that
mean of
999.8, 999.8.
p. e.
69.
An Approximate Method
of Weighting.
show the kind
A
of
work an instrument is capable of doing under favorable conand if work is done only when the conditions are
favorable, the p. e. derived from a certain number of results
ditions
was considered
angle the same
it
weight.
combined
large,
combined
results
unsteady."
"
termination] was
*
finally fixed at
about
60,
it
being found
p. 354.
WEIGHTING OF OBSERVATIONS.
121
is Present.
where the error of observation can be separated into two parts, one of which is
due to accidental causes and the other to causes which are
70.
The preceding
The
total error e
in
the
Let
//,
the m.
s. e. of the
observation arising from the
accidental causes,
the error peculiar to the observation arising from
the constant causes.
^=
Then
/./,,
vation
^,
may
If
m.
//
s.
observations have
e.
fjL g
s. e.
[j.
of obser-
= /V + /V
been made we shall have for the
//,,
is
constant,
when
;/ is
large ./ becomes the important
case
the value of // and consequent
any
little improved by increasing the numbe
but
can
weight
ber of observations.
For the purpose of finding the value of the m. s. e.
It is
evident that
in
observations
is
*
in
general necessary.
Kffort, Su >";)
<>/
///<
122
in
puted, and each of these computed values may be considered an observed value. The values of the declinations
3 are taken from a catalogue of stars.
The errors of 3 are,
of
of
those
and
are constant for the
therefore, independent
,
same pair
of stars.
The
latitude
<p
is
given by
Let
fa
fji s
= the m.
= the m.
s. e.
('
for
pair,
s. e.
the m.
/4j,
of
s.
<p
from one
pair,
and
The
quantity
tions of the
//
will
same pair of
stars, as the
error in declination
WEIGHTING OF OBSERVATIONS.
123
124
stars
is
QV]
where n' is the number of different pairs of
and m' is the number of places occupied.
Now, /2s is found from
and
is,
value
stars
observed
therefore,
may
will
This
be as
/V
An example of a similar kind is afforded in finding the
weights of the angles measured with a theodolite in a triangulation where more rigid values are required than would
be found by Art. 69. The actual error of a measured value
of an angle arises from two main sources, errors of graduaThe former are constant
tion and errors of observation
for different parts of the limb read on, and correspond to
the declination errors above, while the latter are incapable
of classification, and are, therefore, assumed to be accidental.
periodic errors of graduation are supposed to have
The
WEIGHTING OF OBSERVATIONS.
been eliminated by proper shiftings of the
sultant m.s.
and the m.
the
e.
IJL
of a single
s. e. /*
same part
measurement
mean
of the
125
The
circle.
is
re-
found from
of n measurements
made on
from
of the limb
-=*+=
,
e.
!+?
of graduation
of treating this
The method
respectively.
similar to that of the preceding;
and observation
problem
is
quite
is
found.
The foregoing
in
the
observation
servations
is
made with
mean
of
;/
ob-
is
/=<"'+"'
/v
Experience has shown that we
and therefore
it
+ -%may
safely
assume
follows that
in
one
might
fairly
and the
be
rea-
126
where.
(Art. 54.)
So far we
71. Assignment of Weight Arbitrarily.
have deduced the combining weights from the observed
values themselves, or from them in connection with a
But this may not always be
special series of observations.
the
of
The observations may
finding
weights.
way
the best
As we have no
treated by the method of least squares.
direct means of eliminating this kind of error, we must do so
WEIGHTING OF OBSERVATIONS.
27
we
indirectly as best
observation."
72.
It is
we have two
a result
128
a quantity
is
as, for
is
a function of the
observation a weight |,
2 observations a weight
3 to 8 observations a weight
o
1
**
i,
2.
much nearer
has increased.
measured
in 1872
the discordances.
At
* G. T.
vol.
ii.
WEIGHTING OF OBSERVATIONS.
justified
base.
129
in
When
the law of
its
it.
Hence
still
have had
assigned to
it
"
1877, p. 95):
American
As
a third illustration
is
now superseded by
the
we may consider
the weights to
be assigned to a system of differences of longitudes in which
the connections of the stations occupied are interlaced as in
a triangulation net, and the whole svstem is to be adjusted
so as to
If
mav
be computed
in the adjust-
tnen the m.
s. e.
of each
determination
130
s.
as the squares of
e.
But if this has not been done, if in the older work instruments, observers, and methods were poorer than later
and the two have to be combined in the adjustment, the
must estimate as best he can their relative
in a system in
Thus
weights.
Germany, France, and Austria reduced by Dr. Albrecht* the observations were made
between the years 1863 and 1876. The methods of obser-
computer
much improved
in
this interval.
In as-
Weight
Weight
Weight
No change
Change
tions
ot
and so on.
Similarly Dr. Bruhns
Gradmessung,
Appendix
1880.
in
1880,
6.
General Remarks.
is
Astronomische Nachricliten,
2132.
REJECTION OF OBSERVATIONS.
least
is
131
an average ob-
at least
server.
NOTE
II.
way
as in his best
probability
judgment
he supreme, or only
iS
will
give a result of
leading-strings?
as
maximum
computer
is
Various answers
132
we look
may
at
it
Irom one
of error was founded there were no discontinuous observations taken into account.
made with
observations
occurring
in
ordinary work
we
that
section of the observations only
instrument and by the skilful observer.
following from
it
shown
differs ordinarily
made with
the
good
but
little
satisfactory.
A common
76.
ject
is
contained
[of the
summary method
in
angles]
instances
many
by rejecting what would appear bad observations but the rule has been never to reject any unless the
;
REJECTION OF OBSERVATIONS.
133
that he can discover, may make a distinction between measures and mistakes which will do for the set before him.
With observations of another kind he might have a different
mode of procedure. Another computer might have different rules altogether, precisely as in the case of weighting
as explained in Arts. 71-74.
different weight.
may happen,
longitude work it may happen that one of the time stars may
give a clock correction differing, say, one second of time
from that given by fifteen or twentv others observed on the
same night. Instead of rejecting the single star it would
judgment, may
proba-
134
and seek to establish a test or criterion for the rejection of observations which will serve for all kinds of observations.
Of the criterions which have been proposed the
It is as follows: "Observaearliest is due to Prof. Peirce.
bilities
servation
is
"
p. 565.
We
'~
r
vol.
ii.
/,'A*
'"
n-nO(f)
expresses the
the limit
a.
= n\i-
8(t}\
number
But
if
that an error of the magnitude a will have a greater probability against it than for it, and may, therefore, be rejected.
The limit of rejection of a single doubtful observation is, therefore,
=
2n
REJECTION OF OBSERVATIONS.
13$
and registering n observations of a given class. The probability, therefore, is that any record of his of this class of obFrom the average discordservations as a mistake is -n
.
down
mated
in Art. 50,
esti-
about
five
times the
p. e.
mulas.
Nor does itIndeed, most likely it will not be.
follow that the criterion adopted in any special series is of
In
universal application or will receive universal assent.
founded on a general grasp of all the circumstances connected with the making of the observations.
In like man-
136
ner,
and on
rejection.
is
same feeling he
There is no uniform
1
this
will
useful in the
apply
out, in
open
for
improvement.
The
We
sumed
the arithmetic
mean
to give the
We
observations.
"
* De
Morgan (Encyc. Metrop., Theory of Probability," p. 456) suggested that the combining weights might be found from the observed values themselves, but he did not develop his plan,
and
it is
apparently fruitless.
REJECTION OF OBSERVATIONS.
137
how much
tion
in the
introduction of criterions
squares what
When
comes
in.
138
Throughout
to be preferred.
pp. 161 seq.; Airy in do., vol. iv. pp 137, 138; Winlock in do., vol. iv. pp. 145-147 Stone, Month. Not. Roy.
Astron. Soc., vol. xxviii. pp. 165 seq., vol. xxxiv. pp. 9 seq.
vol.
ii.
in
Mem.
Roy.
CHAPTER
IV.
done,
we proceed
independent unknowns X,
V,
^r+/,,F+.
a,X+b, F+
where
#,.
for each
/;,,
/,,,
Ln
observation, and
n { in
number
-L^M> +
- L, = J\L +
(n
>
;/,-),
be
v,
,
(i)
errors of observation.
In practice the labor of
handling these equations will be
as by
trial, for
x-x^x, r- Y
19
=j>,...
140
where
,r,
y,
a,x
-\-
b,y
/,
=
=
v,
v,
(2)
where
+
and
known
are, therefore,
It
is
a^x -\-b.y-\-
always keeping
in
Now,
it
z,
=/,
=4
(3)
=4
the values of y,
if
b t y 4-
+ b y -f
mind
-L -M
<***
quantities.
more convenient
"S +
x=
may
/
as in
is
(2).
be written
'
a^x
//
anx
ln
(4)
'
shown
has been
The weights
mean
to be as a*, a*,
an\
We
I'
a^
'
JL.
an
have, therefore,
_[aT\
-
'
-L, JL,
INDIRECT OBSERVATIONS.
by putting
or,
_ {al}
14!
their values,
\ab~\y
\ac~\z
=j
[aa]
the values of x, z,
Similarly
the
most
known,
probable value of y
if
are supposed
is
_\bl-\-\ba\ X -\bc-\z-.
[W]
and so on
for
^,
to
be
found from
.
of
x, y,
[aa]
_ [/] - \ba~\x -
\bc~\z
- ...
~\bbT
that
is,
[>*>+[^]j,+
[*/]
(5)
which equations are equal in number to the number of unknowns. They are called normal equations, or, better, final
We have thus found the most probable values of
equations.
unknowns
the
in a series of
the mean.*
*
due
to Richelot,
is
worthy
of
>
be
notice.
.r,
y which
2,
satisfy
them
[*/]
best value of
a-
must be that
in
2)
best.
[*]
The
/(-,,
X-
then
if
[**]
[*]
is
142
The preceding
perhaps show
result
be derived
may
in a
manner which
more
still
observation equations,
+ bj
+ b,y =
a*x -f y =
a,x
/,
a,x
I,
b,
all
/3
possible
= bj, bj,
- bj,
a.b,}x = bj,
a b^x
(a&
(bill
bil^a^bi
a^i)
to
is
a 3 i>i)
bil 3 )(aib 3
(b 3 l\
(b 3 Zi
b^l3 }(a^b 3
a 3 b^)
'
[aajbb]- [abjab
Similarly
_[
fl
/][^]-MM
\adbb~\-\ab\ab~\
happens when \kd\
a condition exists
is
as great as possible.
/t's of at least
among the
If
we now
compute the
maximum
[bk]
we
[M] =
\aa\\bb~\
\ab~\\aK\
which agrees with the value resulting from the normal equations
5.
xvi.)
INDIRECT OBSERVATIONS.
143
\aa\x
\bd\x
Hence
i^
>
when
2.
The
residual errors
2
<',
that
is,
+ +
+ v* = a
''/
mill.
we must make
+
Now,
must
+
.
aKX
.
+b y+
n
-/ =a
mn.
dif-
Hence
..-
...
c\s
+
+
in
(6)
each equation,
=[<!
.
=o
144
.,
same
the
is
number
as the
of
f.
.'.
will result.
case.
may be
written
=o
r>]=
(8)
First solution
(a)
Ogden
Direct determination
weight
"
4301
Indirect determination 4326
.'.
(b)
weighted mean
4309
feet.
Cheyenne
Direct determination
6075 weight
"
Indirect determination 6050
.'.
Second solution
Let X,
weighted mean
6067
feet.
Then
(^- Y+
INDIRECT OBSERVATIONS.
Differentiate with respect to X,
Fin
145
succession, and
2X- 7=2552
-X + 2^=7824
.'
X 4309 feet.
Y= 6067 feet.
If
we have
V
with
[/t'f]
^=a min.
/
?\,
Vpv ...
n
_,
in
mum
the
mini
...--= [pbl]
from which
The
,r, y, . . .
relations for
those of Eq.
6,
may
(2)
be found.
(3)
selves.
146
81,
may
be stated as
fol-
results.
in
The second is suggested by the complete form of the normal equations as given in equations 7, Art. 81. Accordetc.,
we compute
equations.
The
in the horizontal
shortening of the numerical work in computing these quanThus with three unknowns, x, y, z, all the unlike
tities.
coefficients are contained in
4- \_ad\x
= [/]
>==[*/]
With
1+2+3+
of computation
.+(*-
!)
amounts
= *(*-
to
I)
quantities.
same way
of weight.
INDIRECT OBSERVATIONS.
147
Ex.
Given the-observation
i.
x
x +y
x y+
=
=3
z = 2
i
xy+z=l
show
that the
=5
+y
Ex. 2. The expansions .Vi, x-2 ,v 3 x 4 for i Fahr. of four standards of length
were found by special experiment to be connected by the following relations
at a temperature of 62 Fahr. (//
one micron.)
,
rj
-l-viv.
+
+
+
find their
iJ-
=
=
-Vi
-r 2
5.932
-v 3
1.0937^3
4-v a
a,
weight
"
39-945
i
1
"
5-371
= 0.006
--v, = - 1.335
= + 14.833
"
"
"
.v 4
4
8
3
6.r 4
7.r,
+
-
72.000 A-O
I2.r 4
8.750^-3
8.750^2 +
13.569.1-3
+
and
x,
= 39.913,
An example
.r 2
will
= 5-932,
now
-v 3
=
=
5-45>
(_).r 4
-V4
+
+
+
128.943
84.993
25.075]
observed
132^1,
to
April
20
1870. at 11
78.940
21.432
148
Star.
INDIRECT OBSERVATIONS.
angles
Vi,
v-i,
in
M^
order by M\,
^ 5 we have
...
and
M-,,
their
149
X' +jc=
= Mi + vi
= J/ + v*
=M + V
Z= M< + v
Y
+y =
Z
Z' + Z=
X' x+Z' + z = X +
- V -y + Z' + Z--V +
Y'
Z=M
v*
For simplicity
+x
+y
+
+
y +
Hence
the
0.76
1.66
iiy
we
x=
=
o".O5,
t/ 2
o".os,
z/i,
72
5.32
42=
6.64
o".36
o".36, ^ 3
2=+o".68
=+o".6S,
z' 4
o".O3,
vb
o".62,
62
Ex.
"
7' 5
find
c' 2
28
59' 40".
n'
64
1 00
symbols
7/4
+ 152= +11.96
47
7-r
We
"7
"4
"
normal equations
z>i
=
=
izx
Hence
weight
z/i
=v*
=v*
34". 56
20' 29 ".So
37
20' 49". 52
36
08'
55". 24
v-,
x, y, z at all.
5.
If the
unknown x
occurs
x + b\y +
x + l>iy +
fi z
c? z
in
+
+
=A
=
weight
l-i
"
i
dz-\-
k->y
4-
Ci z
...=/,
.
=/
weight
150
[For the normal equations found from the first set after eliminating x are the
same as the normal equations formed from the second set directly.]
Ex.
6.
ax + by +
we may
cz
=I
weight
write
+...=/
qax + qby
weigh t
a
_,
+ + +
b,
c,
+ =
/,
s,
.+[/?]=[&]
in
individual tastes.
Thus
[as], \_bs\,
\ls], [//].
may
INDIRECT OBSERVATIONS.
151
figures, it is convenient to have a fixed form for the computation of the terms of the normal equations.
It lightens the
labor
much
cision of the
either in forming, solving, or in finding the preunknowns from these equations, if the computa-
tion
The latter
multiplications and divisions.
With
use in computations of this kind.
of the greatest
the drudgery of
computation is in great measure got rid of. On the Lake
Survey two forms of machine were used, the Grant and the
Thomas.
is
it
series of trials
either ma-
much
made
it
be
but
Form
products
to be
|<7<?|,
With
(a).
aa, ab,
done
[tf/>l,
is
.
to \\rite
them
With
in
Let
us, for
2.1- -|-
+ 3.0*
o.
2y
-j-
0.9 =n
7-,
2.n-f- 1.1=1',
152
INDIRECT OBSERVATIONS.
153
row,
log
and so on
log aj)v
a,a,,
log an
till
is
.,
addition
we
ajv log a^
reached.
By
log
find
\aa],
[ab~\,
[>/],
DW]
Proceed
.
log
log bH
found
log cv
c lt
;
in
and so on
logo
log
cn
till
154
Form
If
(c).
we wish
then since
ab
--%\(a-\-l>y-a-
lf\
and therefore
we form
[fl=itf(* + *)Tsums
[a*]
-[**]}
(i)
the square
/)],
[//],[(/+ *)']
coefficients of y,
a, -\-b
tfo-f-^.,,
3,
., forming the sums <?,-)- /;,,,-{Take out the squares of these numbers
Thus
,,
in the
aa
preceding example,
INDIRECT OBSERVATIONS.
55
saving of 6 entries. The Bessel method has also the advantage that, as we deal with squares, all thought with
regard to sign is done away with. Besides, if the table of
squares is a very extended one, accuracy can be had to a
greater number of decimal places than with an ordinary
As compared with the logarithmic form, then,
log. table.
this method is to be preferred, more especially when the
coefficients are not very different.
On the other hand, if Crelle's tables or a computing
machine is to be had, the direct process explained in (a) is
much
It is
999.v
from which to
By
-(-
find
Check sums.
=
=
O.OOO2J/
3.93
looo.r -(-0.0001^
x and
y.
x' --
oo.i',
placing
i
y'
1004.1101
1 1
4.
1002.9302
o.
o \y
o.i''
9.99.1-'
which are
in
-f-o.oiy =
-f 0.02/ = 3.93
4.
1 1
Check sums.
14.12
13.94
for solution.
156
Solution of
tJie
Normal
Equations.
Before beginning the solution of a series of normal equawe should consider whether the object is to find
(1) the unknowns only, or
(2) the unknowns and their weights
tions
(b)
in
in writing,
are
]j
From
the
first
z,
+[>=[/]
equation
_r
i[ah]l<o
t[tfr]-s-\--[at]-
[aa\
\aa\
(2)
\aa\
+ \bc.\\s =
[/;/.
i]
>=|W.i]
(3)
INDIRECT OBSERVATIONS.
157
where
\bb.i
m -[^[^.1
\aa\
(4)
first of
equations
_[^i]
[W.i]~
in the
3,
[*/.i]
[^.i]
=t^l
(6)
where
(7)
and thence
.r
in (2).
The
first
\CC.2\Z
=\cl.2\
equations.
in the elimi-
158
its
unknown, and
x H- i
a "-
y -f
[<7rt]
\ac\
-
[]
[^<:.
.?
[]
_
=
ij
p.l]"
[^/. i]
[^7i]
In solving a set of
87. Controls of the Solution.
normal equations a control is essential. It is sometimes
recommended to solve the equations arranged in the reverse
order, when, if the work is correct, the same results will be
found as before. But what is wanted in a control is a
means of checking the work at each step, and not at the
;
(b)
The diagonal
equations, and
coefficients \aa\,
\_bb~],
.
...
of the normal
INDIRECT OBSERVATIONS.
For [aa],
Also
\bb. i],
159
tive.
Mc-'Hsj:E] =
':'+
a positive quantity.
Similarly for \cc.2\ [dd.$\ . . .
The principle may be of use as a check in the solution
t
of a series of normal equations which are apparently corThe followrect, but which 'have been improperly formed.
show
this
came up once
in
my own
ex-
7,r+ 77-9.1,2=26
2.3=64
jx
287
12^
127+
-39
9-i.r
+ 77
+ 217
are
9.
= 26
2.9^=38
By
equations
8,
must
(d)
88.
methods of computing
residuals.
[bi>~\
[IT],
is
the
=o
sum
Forms of
Solution.
In applying the
it is
method
of
:6o
arrangement
against mistakes.
Form (a). Solution without logarithms.
The following form has been found by experience to be
It is well fitted for use with the arithmometer
convenient.
For
The computation
The form
unknowns,
divided into sections, each section being formed in a precisely similar way, and in each
section one unknown is eliminated.
Given the normal equations,
x, y,
z.
No.
is
INDIRECT OBSERVATIONS.
From
l6l
Eq. IX.
v
From
-rfr.i]
[*/.i]
,
[^.i]^~[^.i]
Eq. IV.
JiMltW
r
y [**]
[aa]~
\aa\
To
eliminate the
first
[aa']
unknown,
x.
that
is,
In the
first line
write
\aa\ \aa\
first
in
that equation.
first line is now multiplied in order
by [a&], \_ac\,
the
second and fifth lines.
forming
In the third and sixth lines write equations II. and III.
The fourth line is the sum of the second and third, and
The
sum
of the
fifth
and
sixth.
way
The
results
to eliminate y.
value of the last
and proceed
unknown,
,7,
in
the
in a precisely simi-
next results.
Now
first line,
and we have
x.
In order to
We
shall then
[/],
[/>/],
[V/] to
their
proper
signs.
62
(2)
eliminated.
An
efficients of
inspection of them shows that the cofollow the same law as the co-
unknowns
the unknowns
efficients of the
Hence
compute these common terms more than once.
columns.
Thus
[bc.i]
from Eq. VI. may be written down as the first term of Eq.
VIII. This principle is of great use in shortening the
is
large.
it
by the larger coeffiand multiply by the smaller than vice versa. Attento this by a proper arrangement of the coefficients before
cients
tion
in a
considerable saving of
of the
many
places of deci-
ment would
require.
Ex. To make the preceding perfectly plain we shall solve in full the
normal equations formed in Art. 85.
(i) Write the absolute term on the right of the sign of equality, and make
the check sum equal to the sum of the other terms in each horizontal row.
INDIRECT OBSERVATIONS.
(2) Write the constant term on the left of the sign of equality, and form the
check so as to make the sum of the terms in each horizontal line equal to zero.
164
will
values of x,
83.
INDIRECT OBSERVATIONS.
it
[For
is
165
the determinant
[at],
89.
method
is
when
often of service
the
This
number of unknowns
is
large and
simple.
s
\* y\i
If x',y', 3'
i
t ne
z,
and
x = x' -f- x,
y=y' +7,
-"
,_/
'
i
^
*i
normal equations, we
have
C*^,
[*/],
[/] [//],
[#/]
whole, smaller than the original terms [a/],
second approximation will tend to decrease
terms still farther. The approximations are
on the
will be,
[/], [r/].
the absolute
continued
till
Then
x'-^x"-^
.;
yy'-\-y"-\-
...;...
and
the
call
first
[aa]
Substitute for
when
l66
or of z
that
it
is
much
or,
simple cases, the
elimination.
in
method of
ordinary algebraic
substitution,
methods of
in
Braunschweig, 1863.
INDIRECT OBSERVATIONS.
Freeden,
loc. cit.
nutig, p. 129.
Ex. Required
I6 7
to solve the
equations formed
Solution.
in Art. 85.
68
Appendix
8.
equations
+
[>/]
=
+ \bb~\y + \bc\z
\_ad\x -f \ab~\y
\_ab~\x
according to
The
\ac~\s
this form.
function of z.
The
coefficients
(omitting the
Table B. The
terms in lines 2, 4, Table A, beginning with those under z,
are multiplied by [ac], [&M], the coefficients of z in lines i, 3
respectively, and the products set down in lines 4, 5, Table
B. The sum of lines 3, 4, 5, Table B, is written in line 5,
Table A.
Line 6, Table A, gives the value of z.
The next step is to find y and x. The coefficients of
The absothe explicit functions are written in Table C.
lute terms of the explicit functions are written in the first
coefficients of
x and y]
3,
INDIRECT OBSERVATIONS.
line of
Table
169
The value
of z is multiplied by the
Table C, and the products written in
the second line of Table D. The sum of the numbers in
column y gives the value of/ written underneath in line
The value of y is multiplied by the coefficients of y in
3.
Table C, and the products written in the third line of
Table D. The sum of the numbers in column x gives the
D.
coefficients of z
in
value of
x.
The values
of decimals.
now found
of x, y, z are
Denote them by
to
If
three
places
these values
mal equations, and carry out to a sufficient number of decimal places. The residuals are written in the first line of
Table E. The coefficients in line i, Table C, are multiplied
[#/],, and the products written in line 2, Table E.
by
The first reciprocal in Table A is multiplied by the same
residual, and the product written in column x, line I,
Table F. The sum of the numbers in column 2, Table E, is
written underneath, as
The
\bl. ij,.
2, Table C, is multiplied
by
and
the
written
in line 3, Table E.
The
\bl. i],
product
second reciprocal in Table A is multiplied by the same
residual, and the product written in column y, line i, Table F.
The sum of the numbers in column 3, Table E, is written
coefficient
in
line
underneath, as
[V/.2],.
The
Table F.
is
in
multiplied by this
column
z,
line
i,
y=y
/
.
"*
'
I/O
B
Recip.
[W.i]
INDIRECT OBSERVATIONS.
\J\
\_ad~\x
(i)
new unknown.
being the
The
may be
[til]
where
.v, }',
.r',
solution
Form Table C
the
first
reciprocal.
is
(a)
The term
Table C.
\cd~\s'
+ \dd~\w =
The
z.
+ \bd~\y +
now
approximate values of
finished as follows:
is
'^j
The
(2)
[<*/]'
found by multiplying
coefficients of the
new
\_ctd
to
by
equation,
of Table G.
first line
23
\dd.^\, in
is
(a)
is,
-.,
is
\dl'\
of the
an approximate value of w.
new
equation,
This value of
is
(a)
INDIRECT OBSERVATIONS.
seem out of
many,
all
173
Thus Dr.
reason.
him
weeks.
The following
work: Gen.
mentions
that Herr
between the
first of June and the middle of September; and Mr. Doolittle,f of the U. S. Coast Survey, solved 41 normal equations in 5^ days, or 36 working hours.
Baeyer,
in
vii.)
"
vol.
i.
The exceeding
With regard
to a
machine
tAY/,>r/ r.
.V.
Co.tsf
for
performing multiplica-
curofaisrhen Cr,i,hes:siin^.
Surrey, 1878, Appendix S.
iS'.j. p.
109.
1/4
of arithmometer.
For Jacobi's method of elimination, and the applicaof determinants to the solution of normal equations, see
92.
tion
Astron.
1960; Month.
Nachr., 404,
Amer.
Soc.,
Adv. of Science,
1881.
Assoc. for
93.
the
normal equations. If the observation equations are reduced to the same unit of weight, which we shall take to
be unity for convenience, the general form of the normal
is
equations
\ad\x
-(-
Wx +
Let fjt= the m.
/jtx
s.
e.
\ati\y
-\-
\bb-\y
(jL
px
From Art.
,
s.
(i)
of a single observation.
= the m.
of x, y,
=
the
of
x, y,
weights
/,,...
=[_al']
==[*/]
e.
we have
56
P*ti=P3ti = ...=//
In order, therefore, to determine /Jix [Jty
two computations, one of the weights
,
other
(2)
we must make
px py
,
and the
of//,
It is
that
^r,
x
y
in
which
a lt b a
.;,&.
determined.
aj, -f
-f
=M+M +
.
&,
/9 a ,
.
-f aj = [/]
+ /U =
...;...
[/3/]
(3)
are functions of
INDIRECT OBSERVATIONS.
Now, n being
ties J/,,J/,,
which
Art.
differ
the m.
,
M Mv
from
s. e.
lt
Hence
81).
175
since /
/.,,... /
19),
/*"[], ti
H*
= l?[tf~\,
(4)
and therefore
We
b y
a,x
a n x -f- b.j
<**
all
of the
=
=
+ b *y
/.,
I*
same weight.
are
=
=
^ab\-c -\-\bti\y
\ad\x
-f- [afr]
[/]
\bl\
Hence
since
= rrl--,
/,,
/,
(6)
1/6
that
is,
found from
method of substitution.
The general demonstration may be carried out more
simply by the application of the principles of undetermined
Thus
coefficients.
in the
substitute [/],
normal equations
(i),
[/] H- [*][#]+.
=[>/]
.=[*/]
9/]+.
or,
arranging according to
[], + [V^R +.
for x, y,
[/?/],
and
/,,
7.1
/,...
-,}/,
+ [WK + [^R +
I
+
Tlie
unknown
that
the
Hence
[*],
+ WA +
quantities
coefficients of
i;
/,,
/g ,
.,,
may
shall
W+
be so determined
each equal zero.
<
+ [^] +
-f 0/41 +
...
.
a,
o
(8)
same values of
IXfJlRKCT OBSERVATIONS.
1/7
and add
...
a.,,
order,
in
\aa\
[a ft
= o,[ ay = O,
\ca
o, [eft
|=i,
[A?
(y)
/>,,
A,,
|?/}
=
=
o, [cy
o,
i ,
r,,
r..,
.,
etc.,
the second by
.,
Again, multiply the first set by
we
and
have
the sets of
and
an
d
so
add,
on,
flu fl
.
.,,
'/.,,
>
equations
aa -- ab
an
--
'
V.
(10)
y,
is
plain
...
...
|/>/],
94.
[^|,
may be found.
[], [a^J,
same
the
are
these
as those
ot
equations
the
the normal equations, and that
absolute
in
in
terms are
[],
o,
i,
...;
o,
i,
...;...
instead
of [a!
],
Hence,
first
j,
value of
y found from
the weight
of y, and
78
values of [$K],
succession.
unknowns are
[]
INDIRECT OBSERVATIONS.
179
[aa]
x being found
first,
or
where
normal equations
\cc.i\
-\bU\cc\-\bc\bc\
which expressions are easily transformed into
A=
iy
\\hb\-\
.,
ii
....
il>\'
HTTp*
where
/
-\
\l>l>\\<ic\>
- \cc\
180
find the
12
1249
ii
and
12
15
-=
16
in
ii
Ex.
4,
Art. 83.
49
1249
Px
8.4
149
/v
= 1240 = Q
131
If
ux uy
,
96.
ii z
py pz
,
Uy
= O.IO49
Uz
O.IO57
respectively, then
To
INDIRECT OBSERVATIONS.
Perform the elimination exactly as stated
find the values of the
have then
X
unknowns
181
in
in the usual
Art
88,
way.
and
We
82
~\bb.i}^
and
for the
column under
"=irr] =
Also
it is
\_cc.2]
T,
evident that
"
"
r
,
The forms
[],
[/5|5],
[K7],
show
computed from
Thus the sum
is
of
common
occurrence.
In a set
of,
INDIRECT OBSERVATIONS.
Ex.
i.
2A-
+
-
weights of y and
to find the
s.
ix
y4y +
=K
=S
153 = 7'
73
4z
184
97.
the Unknowns.
equations
i,
and so on
Similarly
y=
Hence
since
+ WW\ + [MM +
\ap\\_al-]
[],
[/9/9],
are
this
.,
weights of x, y,
be
stated as follows
may
.
the vahies
of
x, y,
.,
of y, and so on.
coefficients of the
the value
The
lute terms in
of
[a/9],
of the
[7],
unknowns
[/fy],
x, y,
unknowns
\_ac\s
[>/]
INDIRECT OBSERVATIONS.
185
for the
first
un-
known
~
I-,/-.
Comparing
f"xy,"l
AA
[-.2
-r
Eq.
i,
\JC.2]
y~
in
'I
[W.IJ
[<r.2]/
4,
Art. 83.
[I'll
+0.0616
[</]
+ o. 1049
[/;/]
+ 0.0384
[c /]
[/'/]+ 0.1057
[<-/]
0.0224
0.0616
[<*t]
[/]+ 0.0384
Hence
Z'*=
Uy
M*
[] = O.II93
[ft ft}
=0.1049
[?'r]
0.105 7
In
adjusted values
\aa\v
+\ab}y+.
.=[/]
86
tively.
+ \_paV\y +
|>0> + \_pbb -\y-\[paa\x
=
=
[//]
[>*/]
(2)
we should have
x
= [?/][/>/]
[/]
-|-
[>/?][>/]
(3)
and by equating
coefficients
of
/,,
/,,...
in
the
first
expression,
[uaa]a
-(-
[/9]^,
-)-
,,
(4)
To
Single Observation.
is, if
the n observation
equations were
where X y
H
tt
unknowns
we
residuals
i>
equations
rr
tf.
+ b,y -f
-/
v,
(2)
INDIRECT OBSERVATIONS.
where x,
knowns.
y,
We
residuals
are the most probable values of the unmust, therefore, express [JJJ in terms of the
.
order to find
in
i>
From
two
the
A-,
y,
that
it.
sets of equations,
the m.
Now, taking
.
187
is,
s. e. /i x ,
/Jt
to be equal to
in pairs,
by subtracting
...
x
a
to be the errors of
x,
y,
we
Art. 93,
4,
-x
x,
and therefore
J,
= V + //(, 4/[J +
,
/;.
Vtffi
we
;/,-
[^]
Putting [JJJ
;//r,
+ Wr
;
(4)
there results
//-J^in
(5)
;/,-
duce from
i.
2533-
Vn(nnf)
is
known
When
easily de-
(4)
/'=
which
we
,-=
i,
6)
of Computing
[n>].
188
tions,
[?'?>]
is
As checks on
the values of
[_vv\
found in
this
way
the
= o,
[bv~\
o,
.,
we
find
M= -M
its /
we
(c)
If
(d)
We
have
for
-[//]
[*>/]
2[al}x
y,
[rt/]Y
/..,, - [*
+ ([M]
+ LJ/ - L-j)
,
firi]
unknowns
INDIRECT OBSERVATIONS.
Now, from
(9),
189
[aa~],
Hence
'
[.l]
This expression was
\_CC.2~]
The scheme
X
is
Its
as follows:
90
Ex.
in
Ex.
4,
The
i.
To
find the
first
step
is
dicated.
(a)
m.
s.
e.
unknowns found
Art. 83.
to find
[pvv].
in-
INDIRECT OBSERVATIONS.
The solution of the normal equations, with the extra column
added, would be, according to the foregoing scheme,
for [///]
192
Ex.
Show
3.
[Form
that
- x) +
[aa](x
since
[_av~\
= [&v] =
Hence from
Ex.
4.
\ab}(y
-y) +
Art. 99,
3,
- [a/J]
=o
Art. 100
From
the equation
\al}x
\bl]y
+...-[//]=-
and
deduce
^AT.
5.
Prove that
Ex.
6.
From
[<xz>]
= [ySz/] =
\vv]
and
INDIRECT OBSERVATIONS.
193
solution
is
where n F
The value
(6),
Art. 99.
F=f(X,
which A', Y,
observed quantities
in
By
M^
Jlf.2 ,
(i)
functions of the
.
KIn
independently
differentiation
where
are
F,
;/,,
Jt n
J/,,
Mv
are independent,
observed quantities.
of using this general formula
(a) Instead, however,
a great saving of labor to compute
directly, it is, in general,
from a modified form as follows, by means of which much
of the work already done in solving the normal equations
may
be utilized.
to
the
linear
form,
we
have,
IQ4
F=f (*' + *,
or, as
it
may
Y'+y,
be written,
dF=Gs+G.y+...
Now,
since
y,
,r,
(3)
= [/]
[*]*+ [*b+.
we must
where
a^ b v
,,
s,
...;...
/3 1 ,/32 ,
...;...
we may
write
are functions of
Hence, substituting
<7,,
b^
in (3)
[/9,9]
where [],
[;9]
may
(4)
or 97.
Hence
be
written
4
Eq.
(b)
may
/is
F=G& + G&+.
known.
+G.Qn
(5)
INDIRECT OBSERVATIONS.
w here
that
a = [*], +
is,
where
(see Eq.
i,
Art. 97)
+
(7)
Hence
G>+6^+.
of several
independent unknowns
...
x, y,
is
found from
jr=[<2]
where
(2,,
(2 2 ,
satisfy the
equations
z/~
scries
of observation
are required, as
equations the values of the unknowns x, }>,
well as tJieir weights or the weight of any function of them,
these results can be found at one time by making a solution of
.
the normal equations for finding .v, }',... in general terms, and
their numerical values on
then substituting for [W], [/>/],
on the other.
the one hand and the values of G^ G z
,
(c)
This result
may
be stated
in
other forms.
Thus from
Uf
(G1 RA
G?_
~[aa^
[M.i]
G^
(C.fl.+ C.S.
+ g.)'
\cc.2\
Comparing
26
this
196
M
M
M
[*]
[aa\
G-jA',
[fc.1]
(d)
The
expression
(8) for
ii
may
be easily transformed
into
where
substitution.
INDIRECT OBSERVATIONS.
Ex.
i.
In Ex.
4,
Art. 83,
it
is
s. e.
97
of the angle
PSB.
The function
is
dF= -x +
First Solution.
up
From
[aa]
= 0.1193
= o. 1018
Second Solution.
Eq.
4,
2[ay] + [yy]
0.0616
4-
From equations
7,
4(?3=
Hence
(?,=
and
0.0577,
p= - X
i
(?3
0.0577
+ 0.0440
0.0440
= 0.1017
Third Solution. Add the extra column
equations, which would give the scheme
normal
198
Fourth Solution.
.'.
= 0.0833,
(0.0833)o.
12
II/&!
up
=:
I2/
^1=0, k-i
+ (0.0440)^ X
0.0440
9-4625
1016
Also
j.ip= i"-47 Vo.102
Ex.
2.
a?x
+
+
a nx
diX
b^y
b?y
bn
=A
=
I?.
to find the
\aa~\x
+ [ali\y = [>/]
02.
M -f
l
is
From
v,.
the
same
Ii+v
is
pvpv
where
= a * + y+
1
is
v,
(0
t>
M
M -f
Hence
is,
it
if
is the weight of the adjusted value Jlf -\-v^ that
the weight of the function a^x -f- l^y -fand
.,
are the weights of /1? lv
., we have
l
Q,
[>k + [^]^ +
- A
(3)
INDIRECT OBSERVATIONS.
Therefore by substitution of
,,
J5
(?2 ,
lt
+ &j +
199
in (ty,
Similarly
Hence by addition
=
M+
;/,-,
the
This result
[4?]
...
number
may
/,,
terms
;/,.
of independent unknowns.
Hence, since
to
/,...
for/,
[,?/]
and we have
are independent,
as before.
JT.
Ex.
4,
To check
Art. 83.
The weights
of the
The weights
8.4,
Also
7-
4,
9.5,
9.5,
7,
9.8,
<)5
Ex.
i,
Art. 101)
7.5
^+- - + ^ + +
as
it
should.
b.4
9.5
=3
= the
number
9.5
9.h
7.5
of independent
unknowns,
200
From
103.
M M
Hence
spectively.
since
M= V-v
we have
4 = t^i v
-v
=
and, therefore,
[/ J J]
Again, since
where
jj>
is
is
the m.
= [frwr] + [/ mi\
the weight of
s.
e.
Hence
A/V-4"
t*n p *n
-v *=
i
By
'
r> /
*H
addition,
(2)
INDIRECT OBSERVATIONS.
But by
2OI
definition
11
Substituting in
we have
(i),
finally
\-P vv \
I?
71
tti
Miscellaneous Examples
and
Artifices
of Elimination.
problems of imof
artifices
elimination
which may
and
also
certain
portance,
often be employed with advantage in the solution of obserIn this section will be discussed several
vation equations.
104. The labor of solving and finding the values of the
unknowns may often be shortened by taking advantage of
some
principle
inherent in
the
Given
*+^J' =
* + t>*y =
Now,
if
the
first
only,
weight/,
"
t,
whence eliminating
unknown
in the usual
way, we have
x,
normal equation
,
M
~'
r
is
_[
divided by
[/>],
so that
2O2
and from
this equation
succession
is
same
the
as results
y from
for finding
these equations
in
is
the normal
equations.
This process is specially convenient if the original observation equations are numerous and the coefficients l\,
and the terms tlt /,... large and not widely difbv
.
ferent.
The mean
of the equations
in
Ex.
3,
Art. 83.
is
1.97
this
+ O.62.7 =
+ 0.477 =
+ 0.267 =
0.017 =
+ O.OI
-0.07
+ 0.03
0.02
0.257= + o -3
i.o8ji'=
o.oo
+
and
Substitute
105.
for^j' its
value in the
Again,
we may
i.9U'=
'
.
y=
is
0.27
ox .oi4
the problem of some arbitrary quantity to which a convenient value may be assigned.
Thus, to find the difference
of the coefficients of expansion of
INDIRECT OBSERVATIONS.
203
Let x
/,,
/,...
over
at
an arbitrary
We
fa
temperature
/ 1( / 2 ,
/>.,,
-r
x+
(',
(/ a
- =
- /) y - L =
t )j'
<
/,
-,
7-
weight A
weight
<Y) y
\.C
- 0//1
(2)
"AJ
and they
then become
will
O'lJ'
from which
the
If the
.i'
[('
- OXJ
at once,
(4)
from
is
substituted
mal equation
in
in
from which to
find j.
We
2O4
It is
same
y found
in this
way
is
the
For
as before.
OC - ')('- *)] =
EX*
- '.)'] -
The quantity
and
The
/' is
* + (t-*t.)
and
its
m.
s.
The weight
e.
is
//.,.,
is
found from
greatest
when
/ix
is least,
that
is,
when
Ex. The following were among the observations made for the determinabetween the Lake Survey Standard Bar and
Yard and also for the difference between their coefficients of expansion.
The
INDIRECT OBSERVATIONS.
Required the difference of length
ture
/.
Date.
at 62
205
2C>6
7232.9
=V 6^2 =
7-6
7.6
\ 8550.40
Value of
= 831
.r
at
56. 9
5- 1
6.4
= 837.4
62. o
= (i.3)+ (5. i) X
2
// ea
is
(o.i)
1.9
It
probable values.
The
(i)
where a and
tions.
b are constants, to
Also
(X
- *,)' +
F,
-;0* + ...
a min.
x^f
-J-
(aX
-f-
}\}
-(-
(2)
i
minimum.
in (2),
INDIRECT OBSERVATIONS.
207
variables
lt
X.,,
X,
H,
,r a
and
=o
-f a(aX, -f b
Hence a and
/;
straight line is
known.
The problem
little
first
made
use of the
if
the
method of
least
line is
squares.
supposed
drawn the points should deviate as much on the one side as
on the other, and hence that the line would pass through
the centre of gravity of the points.
In topographical
107. Potheiiot's Problem.
three-point problem, usually called
Fig. 10
lows
Let
known
axes
Pa
= c,
and
Jicylragc
~ui
point at which
PJ'P.,
cr, are
Berlin, 1765.
work
the
2O8
observed.
It
is
position of P.
The geometrical
solution
is
simple.
Py
point required.
solution
much used
in practice
as, for example, in
plotting soundings is to lay off on a piece of tracing-linen
two adjacent angles equal to <f> 1} y v and then move this
are
a
figure over the map on which the points 1\, P,
V PP3 lie over the points
plotted until the directions PPlt
The position of
is then pricked
n
v 3 respectively.
PP
P P P
through.
So
in
is
greater precision
necessary.
The
position of
may be computed trigonometrically
as follows: Since the positions of the points
lt
v 3 are
known, the lengths of the lines 1\P P^P^ PJ\ ^re known,
F P P
P^P*
bein
180, we have
known
its
and, therefore,
Call
PP P~ by j, 2 respectively.
of the angles of each of the triangles PPt P3
The sum
can be found.
+ =
- ?, - f -
3 6o
.<
/?
quantity.
Again,
PP,
sin
PP
/>./,
j
sin
.
sin a
.., sin ^
_ P^
- '
PJ
P^_
n_
sin
sin ^,
"
sin
,,
.,
sin
>
tan y suppose,
INDIRECT OBSERVATIONS.
209
and
sin a
-\-
sin a,
sin
sin
-f-
tan
tan
i
,^
y
y
or
cot
I (a,
,)
= tan
(y
+ 45) cot
I (a,
-f
2)
and, therefore,
of
is
.,.
be the known points in order of azimuth, as seen from the unknown point P. X^ Y X^ Y.
the co-ordinates of/3,, /*... referred to some known system of rectangular axes, preferably the parallel and meridian
Let
lt
P.2
X, F, found graphically,
approximations
and ,r, y corrections to these values, so that
to the values of
X=zX'
+ x Y=
t
Y'+y,
Now,
tan
- -
F,
2IO
or
tan(0/
V *d
Y A- v
+ *0 = A - A!
*
-\-
Take
logs, of both
log tan 0/
where o,,
But
+ dfr = log
o2
F '
F,)
+ o j, - log (A-
F'
tan #/
-T77
^r
A,
and
is,
explained
therefore,
A',)
-o
in Art. 7.
known.
Hence
suppose,
or
Similarly
0= *'
values of the
angle P,P1\,
ft
- (^/ -
may
a^x
0/)
=A
be written
(l\
- b^y =
/,
An
enough
in
INDIRECT OBSERVATIONS.
On
211
this
problem consult Bessel, Zach's Monatliche Correspondent, vol. xxvii. p. 222; Gerling, Pothenotsche Aufgabe,
Marburg, 1840 Gauss, Astron. Nachr., vol. i. No. 6 Schott,
;
C. S. Report, 1864,
vol. xii. p. 227.
App.
Ex. Given the rectangular co-ordinates of six known points, and the
angles observed at the point P whose position is to be determined, as follows
(C. S. Report, 1864,
X,
X-,
App.
13):
1845.0
1485.0
F,
= 5534.0
cp
61
12' 10"
212
Hence
48746*
29813*
The
+ 298 13 y =
+ 36767^=
solution gives
y=
which added
to the
3.66
0.02
weight 24572
weight 18533
X= 3451. 66
Substitute for x,
Vi
Hence, since
'Y=
19.5,
\vv\ =
z/2
27.8,
va
8.1,
v4
1402,
,.
_ / 1402 =
4/
.
5-2
22
JLl
x ==
22
ai 4
^24572
22
HY
^18533
final
values
2439. 98
f*
and
177986
109157
= 0.16
i.o,
7/5
13. 5
CHAPTER
V.
We
108.
laid
down
now
in Art. 39.
whether directly observed or functions of the quanobserved, have been independent of one another but
if
they are not independent of one another that is, if they
must satisfy exactly certain relations that exist a priori and
are entirely separate from any relations demanded by obwith,
tities
servation
M Mv
lf
Vn
denote the
observed quantities
respectively.
by the most
satisfied exactly
where
a', a",
a"l
b' , />",
=o
-L'
.../...;
L'
',
are
known
constants.
V^-M^v,
- M, = v,
V,
....
(2)
214
we have
-f-
'V2
b'v,
-f
or
I'
[av\
I'
L'
/'
=: o
I"
=o
=o
where
\_aM~\, I"
V.J/
L"
known quantities.
The most probable system
- \bM\
.,
and
are, there-
fore,
of corrections
is
that
which
makes
\_pvv~\
The problem
is
=a
minimum,
to solve this
GO
suppose.
minimum
function
when
the
Direct Solution
nc values in the
minimum
we should have
function
by equating to zero
unknown
its
n c independent
in the usual way
n c resulting equaThe n
taken in connection with the n c condition equations,
vn
determine the n corrections i\, i\,
Thence \_pvv~]
to each
in succession.
tions,
is
found.
The
CONDITION OBSERVATIONS.
Take
Ex.
i.
Let
Vi, vi,
v3
21 5
measured
PSB +
OSB
Substituting for PSB,
tions may be written
= FSB
= FSB
z/ 4
4- z>5
FSB,
v*
v3
=
=
v 3 + vt
v 3 + z/ s
v\
Vi
vi
vi
their
etc.,
FSP
FSO
0.76
1.66
wiih
Substitute for
SVi
a
-
z/ 4
+ 7~v
?' B
$v i~
7V*
4^3
minimum
in the
+ 4^a + 7(z'i
2
7Vt-
=a
min.
equation, and
+ 0.76) +
2
Va
TI,
iir-a
i.66)
=a
min.
r% v 3 as independent variables, we
5.32
7<' 3
47' 3
4c/ 2
v3
4(1/2
=
=
+ 157-3 =
-
I2-c',
7^1
+ 4rv
6.64
11.96
whence
I/ 8
0".03
0".62
p. 149.
well
+B + C- 180 + e
A
where
e is
(i)
Putting
yl/,
:,,
J/3
-',
for
7- 3
A, B,
C,
the
condition equation
becomes
''i
''i
r- 3
=
=
80 +
/
[,V]
suppose
(2)
Also
"V
Substitute for
7' 3
from
(2) in
7V
the
7' 3
7' 2
+ TV = a
minimum
+ (?, +
min.
function, and
3
7',
/)
=a
min.
2l6
V\
z> a
and
+ 2V* = l
(3)
which give
2,
To
of the
theoretical
an angle, as A.
The function
the difference
is
dF =
'
v\
where G\
i,
and
Qi,
2(?i
@i
that
is,
A =f
weight of
Check.
(b),
if
+
+
=
=o
I
(? 2
2<2-,
of direct
is
unity.
A
(=180+
measure of
Weight
Wt. of indirect meas.
C) of
Weight
of
A =%
mean
as already found.
F-=a
Here
.'.
The weight
is
cot
r-j)
found from
uf
where
dF
sin A
= brsin B
=a
Qi
BQ
Qi
+
+
a s n
'
(?a
"2.Q.I.
cot -^
Hence
Up
2
2
^a'' sin i" (cot
A +
cot 2
cot
cot B)
CONDITION OBSERVATIONS.
Ex. 4. The measured values of the angles of a triangle have the same
weight. Show that if the corrections to the angles are expressed in terms of
the corrections to the log. sines of the angles, and the corrections to these
log. sines found by treating them as observed quantities, the same results
will
be obtained as in Ex. 2.
For example take 50, 60, 70"
Indirect Solution
oo' 30".
MctJiod of Correlates.
unknowns
nc
we
= k'
GO
(|>;]
- /')
and determine
By
k",
.,
we may
"(O] - O +
-f-
k"
k',
k'
3,
Art. 108, in
write
+ O'"] = a min.
(i)
accordingly.
differentiation,
= (a'k + b'k" -f
-f 2p,v )dv
+ (a"k' + b"k" +
1
d<*>
+ 2A^Kz' + ...
2
(2)
we
to find
k', k",
differentials
pression lor </cy, and there will remain n
which are independent of one another. In order that the
;/,.
function
satisfy the condition of a minimum, the coeach of these differentials must be equal to zero.
This gives ;/
equations, which equations, taken in connection with the n c condition equations, give the n unknowns
may
efficients of
;/,.
7'.,,
218
shows that
minimum
equation by
|,
and so write
(i) in
>]
we have
Differentiating,
the form
= a min.
(3)
a"k'+b"k"+
=/X
(4)
(5)
we
Solving,
and
F,,
a,
obtain
'
',
from
k"
.,
and thence
\uaa\k' -f \uab~\k"
where
2
?/
z^
coefficients
85,
2,
from
(4),
written
.
I'
The form
\_uaai],
Art. 108.
(2),
/ / ...
z\,
CONDITION OBSERVATIONS.
2 19
by the
\tiaa\k' -f \iiab~\k"
-f-
-{-...=/'
\iiac\k'"
+ \iibb.\\k + \iibc.i\k'" +
+ \ucc.2\k'" -f
where
/',
/".i, I'". 2,
correspond to
r.\
=/'".2
(7)
respectively.
I'
-~
'' 1
R>
\uaaY
l '"' 2
I
\_ucc. 2\
\ubb.i\
'
,
r
]
R"
8,
Z^e
(8)
\_ucc. 2]
where
._
(9)
/".
i"
29
= I'R +
= i'R" +rs +
/"
.2
/'"
(10)
22O
Ex.
The
Vi
vs
v-t
v3
i,
Art. 109.
+
+
7' 5
=
=
v\
o.
76
1.66
(i)
= 5^1
= iv*
k" = 47/3
= 7^4
k" = 4v
k'
k"
k'
k'
(2)
To form
in
(i),
the
0.5929^'
o. 25/6'
The
v\, Va,
We
from
(2)
find
+ o.25/" =0.76
+ 0.6429^" = 1.66
k"
0.230
2.492
o".os,
The
Check.
v-i
o".36, v a
o".6S,
z/ 4
o".O3, V&
o".62
"<J\
t/z' ]
Hence
When
the
with
The
+ v? + v
same
results as in Ex.
2,
Art. 109.
=a
min
CONDITION OBSERVATIONS.
Note. If a condition equation
of the form
is
This result
is
221
/a
/.
then, proceeding as
to.
Ex. 3. At U. S. Coast Survey station Pine Mt. the following were the
angles observed between the surrounding stations in order of azimuth
:
Jocelyne-Deepwater,
65
n'
52".
500
Deepwater-Deakyne,
Deakyne-Burden,
66
24'
15".
553
87
02'
24". 703
141"
21'
21". 757
Burden-Jocelyne,
sum
weight 3
"
3
"
3
"
Now,
sum
of
measured values
theoretical
359
59'
= 360
oo'
513
oo".ooo
54".
residual error
.'.
Hence, as
sum
in the
".487
preceding example,
correction to each of
first
three angles
-+ +-X
+
i
2".
-3:
".487
.487
X
l
744
is
i",
5",
179
59'
58"
In the primary
S.
Engineers, in
Ex.
with a
5.
"In order
common
several sides.
222
I
discovered that some error had been contracted in taking the dimensions.
it is required to compute the area of this enclosure on the most
probable
Now,
the
to
according
.
ginning,
and departure.
Now,
"
we may remove
Fig.ll
ditions
We
to
therefore,
have,
FE
AF,
fulfil
the
three
other
in
A
the
and
most
following
con-
coincide with
may
these
probable manner.
that
words,
the errors
in
in departure
in latitude
(ai
(rtj
or,
+
+
Xi)
cos (Si
JTi)
sin (3i
+ yi) +
+ ;'i) +
+ jr 2 ) cos
+ * 2 ) sin
(a?
(an
(3 2
(3 2
+
+
+
+
y-i)
y-t)
=o
=o
cos3i xi
sin3i Xi
+ cos3 2
a\
sin3i y\
ai
cos3j yi + sin3 2
a a sin3 2
a a cos3 a
=a
+
+
[a cos
3]
[a sin
3]
=o
=o
'
with
[A*
the weights of
xz
[gy
y\, yi,
.
;
minimum,
being / 1( / 2
q it
spectively.
Hence
k'
i
+ sinSi k"
k' + a cos3i
t
px
l
(2)
k"
q\y\
Tsin
sin
3-
cos
31
T
P*
1L"~7""J~L
sin
cos
f aii
si n
^ cos
3"]
[a
[a sin 3]
cos 3]
cos 3
CONDITION OBSERVATIONS.
Now
if
223
we assume
ments used
in surveying," the
K [a]
[a
k"[a]
from which
The
The
k'
k" are
instru-
cos 3]
[a sin
3]
known.
corrections
common
to
.TI,
cosS, x
a,
sin3i
r,
(2).
cos 31
a^[a r^
[J
cos3 2
in
.r 2
rt.>
departure
to
sin3, xi
stn3 2
<z 2
jr 2
sin3 2
cos3i
sm3
a2
n=
)'_>
v2
cos 3]
w
~
a2
w
r
[a sin
3]
"
rule for balancing a survey:
Say as the sum of all the
each particular distance, so is the whole error in departure
to the correction of the corresponding departure, each correction being so
Hence Bowditch's
distances
is
to
Proceed
in
same way
No. 4 are two solutions one by Bowditch, to whom the prize was awarded,
and the other by Dr. Adrain. Adrain's mode of solution is nearly the same
as by the ordinary Gaussian method.
He employs undetermined multipliers
or correlates, exactly as Gauss subsequently did. To Adrain, therefore, is
first derivation of the exponential law of error, but its first
See Appendix
I.
224
is
101.
The
tion,
m.
s.
first step is to find //, the m. s. e. of a single observaand next the weight, /v, of the function, whence the
e. of the function is given by
To
(a)
find
//.
In Art. 99 it was
equations the m. s.
weight
is
shown
e.
fj.
found from
vv\
where
is
\_pvv\
residuals
z>,
is
the
the
the
;/,.,
and
n(rin
c)
/\pvv\
value
(i)
may
of//.
Checks of
\_pvv\.
When
the
number
in
of residuals
is
large,
\_pvv\
CONDITION OBSERVATIONS.
7-,
7'
=: l/^
Art. no,
4,
a' k'
i/w,
225
be written
may
V k" +
\_uaa~\k'
+
=
-
[/'/]
/'y&'
from
-f
\ubb~\k" k"
Art.
(6),
-f-
+
+ \iicc]k'"k'" +
+
-f 2J>&r]"/r'
...
.
(2)
10.
/"yfe"
\ubb.i\
1 f
_(/7_
/, /,
r,,
(/".I)'
(/-.2)'
[ttbb.l']
\_UCC.2]
n~\
'
'
is
following.
scheme
in Art. 100.
The sum
i.
The
r. ......
/2
_"'
+
M
-11
"M
"*
[]
'
+
.
"
/a
226
Hence
//
Check
(i).
[pvv\
=
=
since n c
[/t/]
~M
as before.
Check
\_pvv\
(2).
pr-y directly
from Eq.
3,
since [uaa]
i.
[u]
To
Ex.
2.
The
first
find the
step
is
m.
s. e.
to find the
CONDITION OBSERVATIONS.
X
= 4.3136
'
[/"'''']
=o.j()
1.2819
1.3395
2.492
To
(b)
find
ii
F=f(V F
2,
lt
and
let
to be found be
is
r)
(4)
it
/,(r,,
ft
ru )=
J\f, n
Expressing F m terms of the observed values, M^M.r
which are independent of one another, and reducing to the
linear form, we have
.
'
where
f)F
f)F
Hence
+v +
as in Art. 101,
?/.,,
?/,,
observed values.
As
in
it
terms of
6M
:,
HL ~" F
*_
o J/,
rtF
oF
i^_.""
()F
oM,
f)
)F
'
i
~"
r)
J/,
228
is
- A
F=a = b sin
sin B
and the condition equation
C= 180 +
A +B+
Hence from Ex.
Art.
2,
no, expressing A,
B = M,
- (Aft
180 +
j-? j
in
+ M, + M,)
Now,
a sin i"
<
\\\
- j^j
COt
cot
A + ~.
cot B\-v\
[]
^ -
[]/
S\
M/
COt
-7' 2
-]
- f#1
W-j
<
COt
/i
+j
L]
Therefore
II
= a'
sin-
A +
cot
rt
sin- i"
-,
(
"-:
cot'
If
is
tip'
and
?/,
jii
r^r
cot
the triangle
s.
1" /<
if
(cot
^ +
cot .5
reduces
this
cot
to
cot B)
equilateral,
[<F-
Also,
cot
e.
= | a" sin
is
\'up
of a single observation.
the weights/i, p*, /> 3 are each equal to unit\r
//
if
jiif
the m.
cot j9
^ +
L] /
and
where
-=
j
a* sin 2 1"
//
s. e. ///,,
would be increased by
in"
and
//*'-
respectively.
CONDITION OBSERVATIONS.
229
It is,
method of
to use the
in practice
correlates.
dfe/X +/',+
This
is
(9)
in
the linear
form,
a ""
a'i\ -f-
/A'
-{-V
=o
/'
~\-
'-.
-/"=o
(10)
with
\
pi>i'\
minimum.
i>.2
.) r,
(/" -a'k'-b'k*-
>,+
(II)
and, therefore,
HF= (f
b'k" --
a'k'
+ (/"
It
.)'"'//,
-a!'k'-b"k"
remains to determine
.)X+-
(12)
Now, when
the most
i' n are
subprobable values of the corrections z/,, v^
stituted in the value of the function dF, this function must
have its most probable value, and, therefore, its maximum
We may, therefore, determine the correlates k
weight.
from the condition that the weight of dF is a maximum
that is, that U F is a minimum.
Differentiate, then, U F with
k'
...
to
as
k",
respect
independent variables, and we have
.
the equations
\uaa\k'
-f-
\iiab\k"
-f-
[uaW+[uMW+
from which k
',
k"
are found.
= \naf~\
=\&rt
(13)
230
= Wf\-\.f\k' -
\ubf\k"
(14)
or
r
it
tip
ff~\
**JJ
-p
-,
-.
...
The form
be found
[#/],
\_ubf~],
are
added
(
v
T
!
c*
\
/
\ubb. i J
[uaa]
as
an extra column
it
ma)'
which
in
the
solution of the correlate normal equations (13), in the manner shown in Art. 100.
For three correlates the scheme
would be
CONDITION OBSERVATIONS.
Ex. 4. To
Here
dF =
vi
.-./,= -!,
From
PSB
7' 3
f-2
=o,
in
Ex.
/ -
i,
231
Art. 109.
=
=
a"" =
a
a"''
.-.
[uaf]
= +
= //"" = +
b"
//"
= X J
0.45
[*/]=-*
[</]
The
= +
0.45
k'
column
for finding
UF
232
Hence from
(15)
M
Also
/<-=
/'
Vu
/r
~~^\V -\uT
The weight
is to
<
SeeEx
1.
'
#l(2 + U 3 )
Ml
the
Ex.
6.
If
Art. 109.
2,
find the
weight
[The solution
The weight
after
adjustment
is
of
/ a>
is
for instance, is
lt
to its
are
n n
:
Ex.
j.
Show
found from
sum
is infinite.
[Sum
180
e , a fixed
.
or otherwise
".
quantity,
m.
s. e.
co
CONDITION OBSERVATIONS.
233
Show
s.
Brest-Greenwich,
Greenwich-Paris,
17 57.154
Brest-Paris,
27 18.190
m.
T
"
s.
7 S7-
weight 14
"
9 21.086
27 18.216
To
9.
weight 10
"
9 21.120
that the
Ex.
de-
12
"
13
free
from error.
Let
Hi,
a-i,
computed from
b ;
whose weight
is
B-i,
\/
measured values
are the
Hi
/>
sin
sin
Bi
Hence by multiplying
a-i
sin
'
'
nn -
sin
being
An
sin fi n
Ai
Bi
sin
sin
We may now
/>,
an
AI
B?
sin
c/i
sin
Ai
sin
sin
An
sin
Bn
(r)
Differentiating directly,
da*
where
Fig. 12
of the angles
(a)
B\/
/A,
required.
If
^
/ \ft
j-
(A),
(/?),
cot
(A)
(B)]
.
v'
\
/
T
i
/ /?
1-^-2
A
i
/ /"*
~r \L*
\
i
A+
cot-
B+
cot
cot
B\
(3)
n/>-
sin- i"
(4)
234
1, i,
Taking
(b)
logs, of both
/log a n
a-^A
i,
members
of Eq.
and
differentiating,
log sin J5 (B
t
(Mi i
= [8A(A)-8 B (B)\
or expanding the
first
da n da H
where S a
is
(5)
member,
=\8A(A)-8 B (B)~\
(6)
number
an
and 8 A, SB
Eq.
in
5,
U loga
and
for
Eq.
= It^V
+ S A $B +
S *~]
6,
/r
and
(b),
8 A SB
SB"]
the logarithmic
is
of the side
in general the
most con-
venient in practice.
Ex.
10.
From
baseAB(=/>) proceeds
~7\t^<
>
all
BA
/ \ /
VA
/
\
B A/
C^'
line.
BN,
Take
first
F = BN =b sin
sin
= 2.
sin
C\
h
Bi
l>
sin
Ai
Bi
cot
Also,
we have
BC,
sin
A?
sin
-B-i
3(83)
sin C*
sin
+ cot C3 (C3 )\
6 sin i"
CONDITION OBSERVATIONS.
235
Hence
1<H =
O.2]
[//J
(cot-
.-/
+4
sin'-
/'''.j
cot- j9,
since
i"
Substituting in Eq.
cot 2 C,
4-
cot'-'
[/J =o
0/.i]=o
=3
[W.i]
[</.2]=0
cot'-
Go"
.-/.j
4-
cot-
+ cot- /^
4-
"
i
.\-
15,
and therefore
where
the m.
is
/i
of an observed angle.
e.
s.
Generally,
dF=(n
4- (w
i) cot,-/,
(A,)
n cot
cot A-,
(//,.)
i)
A (#,) + cot
i)
C, (C,)
cot j9 2 (B.S)
cot B* (/>',) 4- cot
..........
i)
C (C
s
3)
and
If the
since
cliain
J
(iV
'-',
and
^V.\
NN'
is
Hence
follows
it
measured,
at either
it
is
2/>
approximately,
putting
;/
=
,.
sin
yt + io
iSw
4//2/r
-
end.
BN' which
,
and
if
if
then
sn i4
=n AN
jUtfX>
is
'
If
At\"=BA',
we have
in
proceeds
=// A V
is
;/
31
VA
.,
//
as
e.
of
s.
is
XX'
//
AB,
AW
''
sin i"4
/2//
AX-BX,
then
236
Ex.
12.
If
a chain
in.
of
e.
the base
s.
AN,
"
sin
sin
Ai
sin
V4
:i
h b
sin/y 2 sinB 3
sin
is
(n
that the
i)
times
b, is
Fig.14
F= +
tsin^]
sini
BN, show
which
BI s\nB-2 sin
gn-
sin
C-5
sin/? 4 sinj9 5
"1
AB
NN'
is
NN
f.i
n1
'
sin I"A/[
gn
Hence show that in computing a line NN' equal to n times the base AB,
through a chain of equilateral triangles, the least loss of precision is with the
form of Fig. 12.
,
The function
is
sinC
F=b sin^4
sin j?
,
.'.
dF=Fs\n
co\.B (B)
i"|cot^ (A)
colC(C)\
+ (B) + (C) = I
Substituting in (15)
UF
F^ \ ?/i cot 5 A
7/2
cot'
B+
7/ s
__
ZA =Z
If
C,
(T/I
and
cot 2
cot
u^
7/2
7/2
B+
cot
7/3
cot C) 2
= 7/3=-
then
2
HF--
Fsin-
cosec 2
./?
and
Hb sin
/IFwhere//
is the
m.
s.
e.
.,
i"
-
cosec 2
B
-
CONDITION OBSERVATIONS.
E.\. 14.
237
If
ACarc measured
/<=
f^
f.tb
sin i"
cosec-
r>
and
if
KB'
is
would not do
= n,
V3
we solved
If
sin
_
Caution.
/',
the
for
rhombus
it
directly
to take
BB'
b cot
and then form HBB>. The result would be V 2 times too great. For as the
triangles are measured independently, each half of BB' must be considered
separately, so that we must use the form
BB'
( cot
2 \
cot
2 /
+
+
(A)
1
(A
corresponding
to the
+ (C) =
(B') + (C) =
(B)
/,
/,
For the
tk
diagonal
</
y?
cot
4-
c ot
2W cot
h cot
2 / \
where
Fig.16
B\,
are
/>'...,
the
vertical
angles
order.
Now,
/,',
Hence
=.
f =...=
=r
/>':,
422
- cot"
"
all
= B
A'.j,,,
Ji sin
cosec-
similar,
suppose.
in
238
and
tid
2m-\
j
inb~--
{/>"
I
cot 2 '"- 2
Vi6
/>
2
/>'
cot 2 '"" 2
II
cosec
.,
sin-
But
db
If
is
cot'"
Vennessnngtn.
Solution in
112.
In geodetic
work
it
Two
fiber
Groups.
in
due
first
in determining
second group. The
group
to the
* The first
exposition of this method was given by ]>esscl in the Gradmcssiingin Ostfrcusscn.
The method of finding the precision of the adjusted values is due to Andrae, Den Danskc Graii-
Htaaling, vol.
i.
T.andestriangiiliitiitii) vol.
i.,
I!erlin,
will
lie
found
in
Stuttgart, 1878.
minimi
quaii-
CONDITION OBSERVATIONS.
solution
more
239
is
solved simultaneously.
Let the first group of equations be ihe observation equations, u in number and containing ;/ unknowns (;/ > //),
",-i'
+ b,y -f
/+.
- =
/,
weight
c\
/,
"
-l*
v,
b'x-{-b"y-\-
values of the
ID
p.,
/'
in
number,
=o
I"
unknowns
(3)
.r, j',
are
=a
minimum.
(^3)
The value
.
tions, thus
Now, overlooking
and
l>
|C.r)
-f [bb
](;')+
--l/'/i
(5)
240
The
Hence
(x), (j'),
To
are known.
7'.,,
i' n
\ad\xx-\-2\ab~\xy-\-
Ww+.
eliminate
minimum
the
in
by substituting"
which then becomes
T'J,
equation,
2\_al~\x
-2[/>
(7)
[//]
= a rain.
/,
//,... the
of equations
correlates
we
2,
\ni\x
]x
-f-
\aU\ y
+ \bb\y +.
= a'I+
-[l>r\ = a"I+
[al~\
(4)
and
(5),
aa
suppose
2
(8)
or substituting for
(i)
i
|
|,
\2\,
(5),
A'/+B'//-|-C'///
(8),
CV//+ ...
(10)
CONDITION OBSERVATIONS.
241
where
A'
'
15
=
=
-f [>V |//'
(\
now
(i), (2),
It
remains
Substituting for x, y,
condition equations, and
'(!)+" (2)+.
//(!)
.=/.'
"
(2)+.
=//
(12)
where
(i),
(2),
.,
(x], (j')>
from
are
(10) in (12),
[rtA]
//+...
^T//4-.
-\-\_a~\i\
=/.'
~-
where
[rt
A]
(15)
etc.
The
/, //,
of these
solution
.
Also, since
corrections
Hence
j,
the corrections
(V), (j),
,r,
= etc.
(i), (2),
are known.
(6),
the total
242
tions
(i).
*''(!)
'
(2)+.
//(!) + // (2) +
(c)
.=/:
(d)
(e)
The formation
in (a),
|Tj
= *'/+'//+
IT
a"I+b" 11+
(f
They
of the
(8).
(2)
are
and are
The expression
=/'
observation equations
= AV+B*//+.
(c).
substitut-
ing from
(g)
The determination
(e).
CONDITION OBSERVATIONS.
243
First find
(a)
//,
the m.
s. e.
of an observation of weight
unity.
We
m)
have (Art.
number
since
;/
;/
the
is
equations, and
To
//,.
conditions
ot
the
find [vv~\.
Similarl
where
[*7']
=
The
total
sum
TV,,
7',,
[''"''"]
[7-7-]
=O
may
[^']
.
[wzt'J
= O,
suppose.
two
therefore be found in
parts,
one
244
We
from Eq.
and
(i),
,
1
and expand
10,
(2),
;
then
[aw]
m,
or, as in Art.
= /'/+/."//+.
into the
[aw]
Next
into the
14,
form
-- +
p^ + ==, + (^.2)'
(//.I)'
(/')*
for
240.
8, p.
Substitute for
tions S
in a
in
Art. 100.
find the
adjusted values.
dF=gs+g^+.
where ,,,
Put for x,
and
.
Put
}>,
for (i),
are
(2),
(16)
known
quantities.
their values (V)-|-(i),
(j')
2)
and
...
(17)
CONDITION OBSERVATIONS.
where
/,
//,...
4, ...
have, as in Art. in,
.,
we
245
order to eliminate
in
j,
+ O&.+
+/'*,
(18)
We may
determine
efficients of/,
2,
,,
//,...
...
to vanish
so as to cause the
that
is,
co-
so as to satisfy the
equations
DzB^yf
[diX]^ -f
and then we
shall
= [A]
have
JF =&(*)+&(
Substitute for
/', /u ",
from
(13),
and
G&) + ^(7) +
(20)
where
G =g a'k
b'k^
G^g^-a"k,-b"k,-.
l
since
(,r),
(y),
in
(21)
terms of
(-r),
(f),
the problem
is
in
Art. 101.
246
If,
therefore, n F
is
*= [GQ]
(22)
where
Q =[aa-]G
l
the quantities
tions 9.
[],
Putting for
+ [ap]G +.
being as
[/3],
G G
lt
9,
(23)
in the
weight equa-
to (11),
we
find
Si^i-A'^-B'/fc,-
0, = ft -A^ -B^;-.
(24)
where
?i
= [Ui + []&,+
(25)
and
(24),
But from
(11)
and
from
C,, G"9 ,
<2n (2a
their values
(25)
Hence, attending to
(19),
\GG\
or to
\cc.2\
CONDITION OBSERVATIONS.
To compute
by
g gv
where [],
[^7].
247
25, in order,
[/5J,
may
tions.
closer and
+ d'v^
y Vl
+yv +
k'v,
+.
h'v,+h"v
+ k"v, +
.=i
.=/
.
with
v*
=a
=4
min.
248
Let
z//, v,',
solving the
... be
first
the values of v lt
group alone
that
v,,
is,
obtained from
from
= a mn.
If
now
(?>/), (z//),
with
(z/)
of these equations,
solution of the
added
first
found directly.
Similarly if v" v" ... be the values of (v'} obtained by
be the corsolving the second set alone, and (v"), (?'/)
rections to these values resulting from the remaining equations, then since
,
') +
CONDITION OBSERVATIONS.
with
[/(>")']
The
249
= a min.
nv
\,
being positive, the miniwith
the
solution of each set, and
reduced
equation
thus we gradually approach the most probable set of values.
Beginning with the first set a second time, and solving
through again, we should reduce the minimum equation
still farther, and by continuing the process we shall finally
reach the same result as that obtained from the rigorous
mum
is
solution.
In practice the
first
approximation
is
in general
close enough.
It
after
is
CHAPTER
VI.
116.
The adjustment
angulation net
is
shall
preliminary work necessary for the formation of the conIn a triangulation there must be one
dition equations.
measured base at least, as AB. Starting from this base and
measuring the angles CAB, ABC, we may
compute the sides AC, BCbj the ordinary
In plotting the
rules of trigonometry.
can
be
located in but
figure the point C
one way, as only the measurements necessary for this purpose have been made.
Similarly, by measuring the angles CBD, DCB we may
plot the position of the point D, and this can.be done in but
one way. If, however, the observer, while at A, had also
read the angle DAB, then the point D could have been
plotted in two ways, and we should find in almost all cases
F
'
9 ' 17
we
passed.
Thus
the value of
APPLICATION TO TRIANGULATION.
251
ABD
blunt angle
had also been measured we
should have another contradiction, arising from the nonsatisfaction of the relation
If
the
CA
so as to satisfy the
(1)
relations of
The number
stated in the form of an equation in which the most probable values of the measured quantities are the unknowns.
The number
33
number
of
252
of the
method of
solution
is
this:
The form
of the reduction depends on the methods employed in making the observations. These
AOB, BOC
and then
BOC.
B we
B and
pointing at A
the angle AOB.
By
find
now at
next at C, and we have
These two angles are independent of one
Point
the angle
another.
are required.
at
Cm
succession we
If, however, we had pointed at A, B,
should also have found the angles AOB, BOC, but they
would not be independent of one another, as the reading to
119.
reduce,
APPLICATION TO TRIANGULATION.
253
AOC=AOB -\-BOC
must be satisfied between them. By independent angles,
therefore, in the reduction, we mean those measured angles
in terms of which all the measured angles can be expressed
by means of the conditions connecting them.
In the pre-
dependent.
The mean
and
is
free
the pivots.
The
noted
distinction
in the
reverse."
254
Besides those mentioned there are two kinds of systemmeasuring angles that deserve special attenare the errors arising from the regular or
tion.
They
"
"
periodic errors of graduation of the horizontal limb of
the instrument, and the error from the inclination of the
limb itself to the horizon. The effects of the first may be
got rid of by the method of observation, as follows The
reading of the limb on the first signal is changed (usually
after each pair of combined results) by some aliquot part
of the distance, or half-distance, between consecutive microatic error in
scopes in case of two-microscope and three-microscope instruments respectively. Thus if n is the number of pairs of
combined
changes would be
and n
respectively with the instruments mentioned. The operation of reversal in case of a three-micro-
Thus
the
Fig. 18 represent
the positions of the microscopes with telescope direct, the dotted lines show their
if
full lines in
In this
positions with telescope reverse.
the greatest advantage of three micro-
lies
however, of a signal having a high altitude above the horizon, the error from this source may be greater, and then
For an expresspecial care should be taken in levelling.
sion for its influence in any case see Chauvenet's Astronomy,
Vol.
II.
The
Art. 211.
observations should be
made on
at least
two days
APPLICATION TO TRIANGULATION.
255
when conditions are favorable. Results obtained at different hours of a day are of more value than the same number
of results obtained on different days at the same hour of the
This
day.
is
on account of variation
external conditions
in
We
such parts of
finally,
solve
it
after
may belong
to the subject in
forming
in full.
In the triangulation of
the following angles were
Lester, Oneota.
LNO =
256
The second
of these
included
is
in the first,
in general
and the
terms as
An
These relations
Thus,
if
angles, and
Mv
M,,
vv
M denote
M M
n
the
single
measured
or
<',
+ +
?'
*'*
=4
suppose
with
'
The
solution
A denote the
= a minimum
The following
special
frequent occurrence
At
cases are of
a station
the n
i
single
are
measured,
angles AOB, BOC,
and also the sum angle AOL, to find the
adjusted values of the separate angles,
all of the measured values being of the
(i)
same weight.
APPLICATION TO TRIANGULATION.
257
is
or
=/
with
=a
[V*]
suppose,
minimum.
The
that
is,
no)
sum of the single angles, and the sign of the corsum angle is opposite to that of the single angles.
At
(2)
LOA
are
a station
AOB, BOC,
=I
a
with
The
[/#
If the
= -,, = -,.
A
suppose,
= a minimum.
is,
over the
.,
and
[//]
I .
L/J
^
that
solution gives
where
v,
?'
is
l_
258
Ex.
The angles
i.
at station
required to
adjust them.
We have (Art.
120)
?'i
Mi +
vt
z' 3
M-i
Sum
Theoretical
.'.
sum
Local equation
Hence (Ex.
2,
Art.
= 124
= 113
= i22
= 360
= 360
v-i
39' 05". 07
v-t
n'
z> 3
oo'
oo' oo".oo
is
+
weight 2
"
2
+
"
15". 61 +
14
oi".37 + v\ + v? + v
09' 40". 69
i".37
+ v\ + v* +
va
no)
z>z
Z'a
=
=
=
o".64
o".64
o".O9
Check-sum
124
09' 40". 05
113
122
39' 04". 43
= 360
II'
= i".o7
15". 52
oo' oo".oo
we may deduce
+ vt + v 6
at station
South Base
v6
122.
23
08' 05". 13
47
31'
70
39' 2 5 ".04
19". 91
at a Station.
If s sta-
tion in these
equation.
measured
indicated by the
number
of superfluous angles,
s-\-
I.
is
APPLICATION TO TRIANGULATION.
123.
259
With
single
are to be found.
Now, two
will
points of a line joining any two of these points will determine a fourth point, and. so on.
Hence to determine
the s
2 points, 2 (s
are
If, there2) angles
necessary.
fore, ;/ is the total number of angles measured, the number
of superfluous angles, that
be satisfied, is
n
Ex. In a chain of triangles,
number
The equations
into
if
2(s
two
124.
is s
number
of conditions to
2)
s is the
of conditions to be satisfied
number
the
is,
number
;
and
of stations,
in a
of conditions
show
that the
chain of quadrilaterals,
is 2s
4.
-The
sum
of the angles of a
drawn on
a plane surface
is
_area of triangle
R sin
\"
Vp
ft t ,
34
260
question by conforming
puted from the formula
(in
it
mean
a, b
are
two
seconds)
sides
and
tp
of
=
C
C
7
2,o 1
where
of the latitudes
is
/> 2
sin
triangle.
= A ab sin C
when
log
may be
lowing table
is
APPLICATION TO TRIANGULATION.
To
find a,
be
b, tp,
26 1
must
made
first
Ex. In the triangle N. Base, S. Base, Oneota, if 573, #4, v* denote the corrections to the three angles, we have for the most probable values
ONS=.
122
NSO=
23
n'
SON = 34
Sum = 180
Theoretical sum = 180
and the angle equation
is
+
+
39". 66 +
15". 61
z> 3
08' 05". 26
z> 4
40'
z>7
oo' oo".53
oo' oo".o5
= 180+
z 3
z' 4
z- 7
+ v\ +
vi
The
result
is
+ o".4S = o
Similarly from the triangle Lester, S. Base, Oneota, the angle equation
Ve
125.
Z'7
V6
"'a
I*.
10
is
in
a Net.
It is to
be
262
NL
omitted.
is
Generally,
if
s is
the
number
of
stations occupied, the polygon forming the outline of the net will give
to
rise
diagonal that
is
drawn
Hence
in
Each
form a
to an additional
/ lines
angle equation.
the
number
of
in
both directions,
sighted over
diagonals
will be /,
s, and the number of angle equations
if
Z-t+l
of the lines are sighted over in one direction only,
the total number of lines in the figure, then since
If /2
and
/ is
/,=/
/2
the
number
pressed by
equations would
of angle
be
ex-
l-l-s+i
Thus
in
of figures,
LONS
ONS
LONS
ONS
OLS
OLS
+ + + + + 3-06=0
+ ^ + ^+0.48 =
^ + + ^ + v + v + 3.06 = o
V + V -\-V + V,+ 1-10=0
+ + + 0.48 = o
!
'
'.
'.
^3
v*
K
^3
v<
v,
*0
=O
APPLICATION TO TRIANGULATIOX.
263
W4
+ + *'+ 1-37 =
V.+ 1.07=0
+
*'
V.
ABCD,
in
which
all
Fig. 22
in
angle equations.
But
in Fig. 19,
Fig.23
L.,, 7, 3
264
SNL
and SL
sin
SN _ sin SL,N
sin
SL
sin
SN ~~
_ sin SON
sin
SL O
sin
sin
SOL
sin
SL
sin
SNO
Hence
is
to
SLN
sin
SNO
sin 5(97.
sin
SON
sin
called a
.rzV&
is
sTn~52VX
sin
which
ABCD
Di
"^ +
.#CZ> 3 +
"
Fi 9- 24
D,
BD^A =
=
=
+ CAB
CZ> 3J5 + DzBC
-ffC/4
180
180
1
80
+
+
satis-
fied
The
DAB
BDA
sin
sin
BCA
CAB
sin
sin
CDB _
BCD
side equation
sin
SLN
sin
SOL
sin .S7V0
sin
SNL
sin
SZ<9
sin
S6W
SN
sin >SX
Hence
in
50
sin
SL
sin
sin
SO
sin .S/V
forming a side equation we may proceed mechaniway. Write down the scheme
cally in this
SN SL 50_
SL
SO
SN~
APPLICATION TO TRIANGULATION.
265
lines radiating
first
The
point
equation.
is
fixes
Ex.
i.
In a quadrilateral
ABCD, in
which
all
show
that of the 15 side equations that may be formed, 7 only are different in
form, and that by taking the angle equations into account all of them may
one kind.
266
involving quad-
equations as in a quadrilateral.
Thus in the figure which represents part of the triangulation of Lake Erie west of Buf-
falo
Fifl.25
tions from
GHFA
HGDEF
taking
as pole,
GA
GB
GC
GB_
~GC
~GH
GH ~_
"GA
GBA
sin
GCB
sin
sin
GAB
sin
GBC
sin
GHC
GCH
Linear Form.
sin
sin
GAH =
GHA
Thus
far
we have
done.
sin
sin F,
~
where
angles.
F,,
2,
Let
Mv
APPLICATION TO TRIANGULATION.
v
vv
267
(M
-\-
sin
(M,
sin
z/a
sin
-f v t )
(M
(M
+v
'
4)
M, + -JM
log sin
v,
f
which may be written
two forms
in
for
.=o
(3)
computation
expressed
:
seconds,
where
d' is
o'v, -f
that
\"
in a
Then we have
in
we may put
+ log sin M,
is,
[oV]
log sin
M,
-f
=o
(4)
known quantity.
Secondly, we may replace
where
/ is a
by
mod sin
\"
cot J/,
Cot M,
v,
Eq.
cot
then be arranged
may
M, v
-^ 9
10'
if
mod
sin
'
(log sin
M.
log sin
M.
is
chosen as the
unit.
.)
(5)
268
For the method of computing log. sines and log. differences for small angles or for angles near 180, and also if a
ten-place table is used, see Art. 7.
The side equation deduced
130. Check Computation.
from spherical triangles must also follow from the corresponding plane triangles, the angles of each spherical triangle being transformed according to Legendre's theorem
that is, for example, we should obtain the same constant
term / by reducing to the linear form the equation
;
sin
SLN
sin
sin
SNL
sin
SOL
SLO
SNO __
SON~
sin
sin
or the equation
s'm(SLN-^)
sn
where
s 19
sin
s'in(SOL-^)
(SNO -
% ~
_
SON respectively.
to be
if
etf'
- <T) -f f (d'" a
<T") -f
(d"'"
- <5""") = o
or
f,(cot
SLN
cot SNL) -f
fa
(cot
SOL -
(cot
cot
SLO)
SNO -
cot
SON) = o
APPLICATION TO TRIANGULATION.
Ex. The quadrilateral N. Base,
S.
269
19).
Take
We
LS LN LO _
LN LO~LS~
from which we write down the side equation
sin
sin
that
is,
+
(Hf., +
sin
(Mi
sin
7- a
z-s,)
sin
(M +
a
sin (A/i
First
39' 05". 07
log sin
43
46' 26". 40
log sin
70
39'
log sin
47
24". 60
7' fi
sin (M-,
Form of
+
+
+
(M + V
+M + +
sin
7-1)
K)
7-,
Reduction.
=9.9618969,7
7',)
= 9.8399903,4
7- 2
7- 6
7- 5
7- 6 )
9,227-2
=9.9747656,9+
21,98
7- g
7,397-6
530,0
3
+
40". 69 +
06". 06 +
31' 20". 41
=9.8677859,5 + 19,287-5
=9.9177470,2
14,297-!
7-,)
7- 7
z- b
= 9.991
180,3
4,30(7-7
rv)
510,0
Hence
14.297-,
9.227-.J
19.287-5
is
7.397-6
4.307-7
17-687-8
20.0
Check of the constant term by computing the log. sines after deducting
to which it belongs.
\ of the spherical excess of the triangle
"o
39'
-4"- 4S
2/O
for the
LSO.
The
first
Second
Log
Form of
Reduction.
Log
Sin.
Sin.
0.43802/2
9-8677859,5
9.8399903,4+ 1.04372/8
9.9177470,2
9.9618969,7
9.9747656,9
9.9911180,3
0.35102/6
sum
0.91562/5
0.67862/1
0.2043(777
t')
530,0
510,0
io 7
mod
20,0
log I.30IO3
log 8.67664
sin
9.97767
0.43802/2
0.950
is
0.91562/5
0.20432/7
0.35102/0
0.83942/8
0.950
=o
This result may be checked in the same way as in the first form.
In reducing a side equation to the linear form the coefficients of the corrections should be carried out to one place of decimals farther than the absoThis for a short computation would be unnecessary, but in the
lute term.
reduction of an extensive triangulation net it is rendered necessary by the
last figures in
products and
quotients.
It will
we have
Indeed,
it
is
in
general sufficient to
carry them
the labor of forming the log. sines with a lo-place table, and
then cutting down the results to 8 places, is, in very many
may be
used.*
* On the Coast
Survey and Lake Survey the practice in primary triangulation has been to
carry out the log. sines to io places of decimals. On the English Ordnance Survey they were also
carried out to io places, but on the more modern Great Trigonometrical Survey of India to 7
places only. In the triangulation of Denmark, Andra: used 8 places
lowed by the Prussian, Italian, and other modern European surveys.
and
in this
he has been
fol-
APPLICATION TO TRIANGULATION.
We
131.
2/1
are
be
it would
most convenient to
-fput the side equations on the same footing as the angle
equations. To do this we may divide the side equation by
such a number as will make the average value of the co1
or
i, it
efficients
with
follows that
equal to unity.
in triangulation,
would be
1.43^-0.92?',- 1.93^
+ 0.74^-0.43^ +
i.777/ 8
+2.00 =
it
may be
(Compare Art.
58.)
of Pole.
2/2
is
notice.
at
introduces the smallest number of unknowns into the equation, for then the normal equations would be more easily
formed.
If the
ployed
is
em-
inter-
it is
taken as a base,
point
its
we must know
To
fix
a third
two
the other
2)
that
is,
2^3
lines in all.
number
is
/,
the
num-
APPLICATION TO TRIANGULATION.
her of side equations, as indicated by the
fluous lines, is
273
number
of super-
/-2.J+3.
134. Check of the Total Number of Conditions.
Leaving
local equations out of account, if/ is the number of lines in
a figure sighted over in both directions, and s the number
s.
of stations, the total number of angles in the figure is 2/
is
/-2J+3
.*.
the total
number
2/
where
But we
;/
is
/2
of condition equations
35
-f- 4,
that
is,
11
is
2s
-{-
the number
conditions to be satisfied
among
n
2s
number
-f-
of
is
We
Manner of
In the selection of the angle and side equations in a triangulation net we have two dangers to guard
Equations.
against:
first,
second, that
and,
rule
For example,
Fi s- 26
Fig. 27 represent a trias base (Fig. 26), a third station,
Taking
let
AB
angulation net.
an
C, gives
angle equation from the triangle
ABC,
274
ABD,ACD;
from
ADE ;
ABE,
side
equations from
fifth point,
angle equations
equation from
ABDE.
We
have thus
(Fig. 27)
Fig.27
in
tions.
Number
Number
.'. Number
and Number
Check.
As an
of stations
of lines
=
=9
5
of angle equations
of side equations
of the
illustration
=9
2x5+3 =
dif-
Fig.28
(1877).
From
down
in
The
is
DE.
Thus the
DEF
EDF
00
oo'
FED =
00
oo'
DFE=i 79
59'
179
1
80
oo".8i5+^
i".
854-1;,
5 6".733
+ +
*,
59' 5*
oo' oo".ooo
o=
i".267 4- v 4- v a 4- v^ 4- v,
t
APPLICATION TO TRIANGULATION.
275
ADF
BDFE, GDFE.
o.oooSv,
54.772 i?,
-f-
o.ooi 5^ 4
Now,
4o.7o66(? a
-f~
^)
+ *0 = 40.
0.0003 ( v
26. 1 803*',
-f-
54.772^
which
is
triangle
+ i.oi8(? +
s
v<)
-j-
? by
v3
+ 0.654^ =
t,
DEF.
AEDF, BDEF
coefficients less
than
give respec-
-f-
Again,
equations
in
a net care
independent.
36
show
276
BDEG,
sin
EBG
sin
sin
sin
~sm
sin
sn
sin
sin
sn
sin
DBG
sin
sin
is
common
BAG _
ABG
AEG
to
all,
we
sin -ZL46P
AEG
sn
EDG
sin
BEG
to the identical
form
i
come
indeterminate.
the rule given on p. 273 is followed closely this
repetition of conditions will hardly occur.
If
Fig.30
Fig. 32
Fig.33
Fig. 34
APPLICATION TO TRIANGULATION.
277
Side equations.
Angle equations.
From
Fig. 29,
From
Fig. 30 in addition,
Triang.
AGB
AEB
BEG
Quad.AGSE
From
DAG
BDG
AGBD
"
AGDE
"
AGDF
"
BDFG
"
KFEG
"
ACBE
CBDE
DAE
Fig. 31 in addition,
From
"
DAF
Fig. 32 in addition,
BDF
From
BFE
Fig. 33 in addition,
From
BAG
Fig. 34 in addition,
"
In
be
all
10 angle equations
and
(p. 274).
138.
We
See
Thomas
for
an example Verification
Maclear,
vol.
i.
and Extension of La
'
Cailles
Arc of Meridian, by
Sir
278
perience shows that the solution of a series of normal equations is much facilitated if the coefficients are arranged as
the steps of a stair rather than irregularly.
Thus
= [al~\
= \bl]
\ab~\x + \bU\y -f \bc~\s
\ad\x-\-\aU\y
\bc\y
is
+\cc-\z=\cl}
[ab] y
-f-
[/]
therefore be so arranged
cannot always be done at the
first trial), the normal equations will fall in the first of the
above forms rather than the second. A good rule is to
begin with an angle equation, proceeding from triangle to
triangle until the points gone over are covered by a side
Continue the process with
equation, and then introduce it.
that, as far as possible (and
it
Fig. 35
where angles so
small
occur, to avoid trivery
angles involving angles immediately contiguous to these small
in Fig. 28, and,
angles.
139. In the explanation of the
we have assumed
that, the
com-
assumed
routes
we proceed through
as
chains
base,
of
by
different
The
may make it
triangles.
necessary to proceed through polygons, and then the formation of the side equations becomes more complicated.
APPLICATION TO TRIANGULATION.
A good
illustration
occurs
in
279
DG
quadrilaterals
This new
ECD=}>.
We
equation.
From
the
take the
first
CD
line
pentagon,
ABCED,
CE CD CA CB__
CB CE~
~CD CA
or
sin
(EDG -fsin
sin
-*")
CED
sin
CAD
(ADG -\-
CBA
sin
sm CAB
sin
sin
x)
DEGFC,
CEB =_
CBE
pole at E,
EG EF EC
EG EF EC ED
^
or
sin
sin
EGD
EDG
Now,j can
CDE; thus
sin
EFG
sin
sin
EGF
sin
ECF
EFC
sin
(EDG-\-x] _
sin,)'
(3)
where
f is
1872.
CDE.
280
Eliminating
x from equations
i, 2, 3,
To
cot
find
it
write
(i) in
the form
(EDO -f x)
sin
sin
ADE
sin
ABC
sin BAC
CAD
CED
sin
sin
EEC
T-,
cot
sin
ADE
140.
Ex.
(Fig. 19).
The method of forming the condition equations having now been exwe are ready to adjust the quadrilateral NSOL, as promised in
plained,
Art. 120.
Collecting them,
all
been formed
in the
preceding sections.
i, 2,
Art. 121)
v\
v?.
w4
4-
v*,
z> 3
v6
=
=
1-37
1.07
Art. 124)
Vs
V6
+ V4 + V^
+ VT + V + Vu =
S
0.48
1.
10
Side equation, the unit being the sixth place of decimals (Ex. Art.
1.43^1
The methods
0.92^2
i.
93^6
+ 0.747/0
0.43^7
i.>7z/ 8
130),
2.00
of solution have been explained in Chap. V., and we shall proceed in the order there given for the four forms.
APPLICATION TO TRIANGULATION.
FIRST SOLUTION
METHOD
281
OF INDEPENDENT UNKNOWNS.
there
Expressing all of the unknowns in terms of these four, we write the equations
in the form of observation equations, as follows (see Art. 109):
Va
weight
"
7'3=
=-
-i-37
Vi
7'4
i-
+
7-B
7-3
=
=
Hence
7-1
0.5657',
0.435T-1
7-4
0.6617-4
37' 2
37/2
"
7/4
7-o
7' 4
7-2
O.66l7' 4
"
+
+
1.07
0.89
0.672775
I.36l
r.6727'5
1.699
14
23
6
282
To
(a)
find the
m.
s.
e.
of an obser-
[pw]
7-
53
the
extra
Hence
9-4
i".23
(b)
angle.
F=NLS
=
.'.
Hence from
180
C/f=
772
(M-i
+ v* + Mt, + v
6)
7^6
the extra column, the sixth, carried through the solution of the
284),
UF
= 0.053
and therefore
Up(c)
To
find
the weight
1.23 1/0.053
and m.
s. e.
base,
to
F= OL
~_
sin
S'JL^^
sin
sin
{Mi +
in
Lso
OLS
VT) sin
two ways.
(M + v)
APPLICATION TO TRIANGULATION.
/'directly; then
Expand
(i)
283
6F
dF= / TTT
\dMs
6F
+ nrr v
dM*
*'
0.05057/3
0.0077/1
+0.1717-2
6F
TUT
^M^
6F
''>
+ ^nrr
dMa
O.Il607-7
0.02827'e
\
T
'
sin
0.13427-9
+0.2537/5
0.0567/4
re-
UF
= 0.0019
Hence
I.I
F =1.23 V0.0019
= o'".o5
(2)
log
Take
members
logs, of both
/'log NS +
log sin
(M +
3
v3 )
of the equation
log sin
(M +
is
f/log
Hence from
constant,
1.337-3
0.747-8
0.187-1
4.5O7' 2
F=
z> 6 )
v-,)
column addejd
3-04^7
from equations
=
Now,
23
and
F= dF
mod
t
1
(p. 255)
F= 16556^,
.-.
37
f.i
om
.o6.
r.
normal equations,
=I
+ v
3-527'a
1.457/4 4- 6.637-5
Also,
^togF
F=
the last
we
then
place of decimals.
284
The
-k
APPLICATION TO TRIANGULATION.
The Correlate Equations.
I.
232/4
286
The
Precision.
From
To find the m. s. e.
we find directly
(a)
[pvv]
7-53
tions
respectively.
Hence taking
being
the
mean, [/w]
of Art.
= 7-54,
5,
".23 as before.
From
/<
the values of v
the values of u,
the values of
f given
a,
by
[ucf]
=
/] =
[uff] =
0.782
[<(/"]
0.167
0.500
U
[
o.
\ubf~\
the
e.
in
connection with
this function,
[naf]
Hence from
s.
o.
+0.667
(page 287),
uF
= 0.053
= 1.23
= 0".2S
and
fi
Compare Ex. 4,
(c) To find
Art.
o.o53
in.
the weight
side, the
dF~=
0.05057/3
0.02822/6
0.1160277
0.13422/9
=
[/] =
[ucf] =
[uaf]
0.0046
0.0036
0.0073
["/]
[*/]
[//]
=
=
=+
0.0040
0.0165
0.0030
of a
APPLICATION TO TRIANGULATION.
Hence from
the eighth
column
UP
and
normal equations,
0.0023
finally,
HP
1.23 V.OO23
= o".o6
Solution of the Normal Equations.
I.
287
288
SOLUTION
IN
Two
At North Base.
GROUPS.
APPLICATION TO TRIANGULATION.
(b)
At South Base.
290
(a)
SNO
NSO
"
NOS
=
Sum
122
180
oo' oo".3i
180
oo' oo'.os
Angle
"
180
n'
15". 52
(2)
(i)
+ (4)
39". 66 + (7)
23
08' 05". 13
34
40'
o=o".26-(i)-( 2 ) +
(b)
NSO
Angle
SOL
OLS
70
39' 25". 04
78
27' 06". 06
30
53' 30". 81
+
+
180
oo' oi".9i
180
oo' oo".37
S.
sin
LNS
LNO
LNS = 113
39' 04". 43
(2)
LSO
70
39'
25". 04
(4)
LON=
43
46' 26". 40
(8)
9.9618975,6
9,22
9.9747660,1+
7,391(4)
539.1
509,4
29,7
(7)
(4)
(5)
(7)
(8)
(9)
(7)
+ (8) + (9)
sin
9.8399903,4
4)
Base.
(c)
sinLSO
sin
NSL
sin
sin
LOS
LNO
+
(5)
124
NSL =
LOS =
(S)}
09' 40". 05
19". 91
(5)
27' 06". 06
(7)
(8)[
47
31'
78
19,28(5)
9.9911180,3+
4,3o|(7)
9.8677849,8
509.4
(i)
14,29(1)
9.9177479,3
(2)
+ 21,98(8)
LOW ~
_
(8)
APPLICATION TO TRIANGULATION.
Check by deducting
J-
29!
angles.
09' 40". 01
39' 04". 36
124
70
39' 24". 92
47
31'
43
46' 26". 36
78
27' 05". 93
113
9.9618976,2
19".
84
9.9177479,9
9-9747659.3
9.8677848,6
9.8399902,5
9.9911179,8
38,0
8,3
8,3
29,7
glance
well.
at the log.
differences for
i"
shows
that
by expressing them
43(1) -0.92(2)
+ 0.74(4)-
19(5) -o.43(7)
1.
is
1-77(8)
2.97
in
We
unity nearly.
=
(2) =
(4) =
(0
+0.267
i
|
0.233
|~2~|
2
|
+0.038
(5)=
0.021
(7)=
j~4~j
0.021
+0.088
j>J
|~5~j
+0.032
(8)=
|~7~|
+i.ooo|T]
(9)=
+0.125
292
in
Terms
of the Correlates.
APPLICATION TO TKIANGULATION.
+
-
+
+
0.92(2)
0.74(4)
i.i'X5)
0.43(7)
1-77(8)
-f-
+
+
+
0.138
4.706
0.089
O.O22
6.418
- 4-93
(collected}.
0.049
0.021
1.241
1.687
=
=
0.260
1.540
2.970
II.
III.
Substitute for
0.804
0.564
o.oSo
3-133
+ 4.706=
The solution
III.
II.
+
+
0.852
0.533
0.038
0.144
0.006
1.687
I.
Check.
III.
II.
293
= -
1.597
=
=
0.394
-0.642
their values in
(4),
rections.
Adding the local corrections and general corrections together, the total
corrections to the measured angles result and are as follows
:
Local.
294
The Precision of
To find the m. s. e.
Computation of [/r^].
of
(a)
(r)
From
an observation of weight
unity.
directly
thus
[>] = 7. 50
(2)
Check
From
(Art. 114).
N. Base gives
S.
Base gives
(p.
3.24
(p. 289)
4.99
From
[vv]
288) 1.75
[z/V].
X
X
X
/'
(<r]
/"
/"'
=
=
=
I.
II.
III.
X - 1-597= +0.42
X 0.642= + 0.99
X 0.394= + 1. 16
-0.26
1-54
2.97
+ 2.57
X f===f=- 0.26 X -1.885=
'
(/5)/
LAj
.1
/o"l
X F =z: = -
+0.49
1.45
1.183= +
-94
0.394
1.72
OB.I]
/o'"2
/
.2
Xr--=^
0.37
2.58
[fC.2]
[ww]
.'.
and
[p-cn>]
=4.99+
2.58
2.58
= 7-57
Hence taking
the
mean
To
find the
m.
s. e.
Angle
...
= NLS
^=_( )_(
2
= + O.O55
From equations
5)
I-
0.067 II-
O.7OO III.
APPLICATION TO TRIANGULATION.
The values
of [anr],
q\
[a/if]
=+0.267 X
0.233
= o. 233 X o + o. 267 X
0.021 X
q = +0.038 X o
=
0.088
0.021
X +
X
q^
q-i
3
I
I
=+0.233
= o. 267
= +0.021
= 0.088
295
Hence
296
I.
APPLICATION TO TRIANGULATION.
[Number
of conditions
=w
i,
where n
is
297
number
of observed dif-
V
7'i
7> 3
T'a
V4
Z>5
Vg
+V^
+ Vt + Vg
V\n + V\\
+ Vio +
7'13
The weights
= + 0.086
+ Vio = + 0.045
7'g
+ Vu
Vn =
Ex.
Koppe
The system
z.
of triangulaiion
the deter-
in
tunnel.*
Goschenen)
Gothard
St.
In the fol-
lowing table the adjusted values are given side by side with
the
measured values.
proposed as a QOSCH
problem of adjust-
It
is
lent.
At Goschenen.
Measured.
shown
0.049
0.096
e.
in the figure
was executed by
298
At VIII.
APPLICATION TO TRIANGULATION.
299
is
second,
groups of condition equations
the
must
be repeated till the reclosely enough,
process
is
attained.
quired accuracy
To make the operation as simple as possible let us take
satisfy the
first,
The
solution
"i\
(2)
is
given
o".64,
solution
is
v =
4
(3)
Ex.
?',
i,
= -
Art. 121,
o".64,
7' 3
o".O9
The
in
given
+ ?'.- + 1.07=0
7',
in
Ex.
o".i3, v,
2,
= -
Art. 121,
o".so,
TV,
+ o"-44
Angle equation,
?' 3
+ + + 0.48 =
?'4
*',
< ;
The method
*',
39
first
+ + + 0.26 =
of solution
o".i3,
?' T
*'<
is
7'
Ex.
2,
o".o8, r.
given
in
Art. no,
o".os
approxi-
successive
approximations
added, give
v6 =
v,= -o".64
V3
Vt
=-O
V7
O".22
ff
found
so
far,
when
-C/.55
C/.05
.2I
Proceed similarly with the remaining two condition equaThe resulting values will agree closely with the
tions.
still
more
.38
The measured
in
the
APPLICATION TO TRIANGULATION.
Station
occupied.
301
3O2
The
Arts.
tit
Each
(a)
(b)
Station.
the table.
(a)
and
(a)
(b),
Sum
we break
we have
If
Angle.
Adjusted.
22". 73
06 ".03
28". 76
28". 75
22 ".65
05 ".96
25"- 35
06". 04
check.
APPLICATION TO TRIANGULATION.
303
144.
we
ment being
be
weight.
into
its
We
We
simplest parts.
quadrilateral
SCWK,
next
have
two
up the net
our figure,
in
single
triangles
first
KWA,
AWD,
Adjustment of a Quadrilateral.
SCKW all
(i)
WKS may
i, 2,
In the quadrilateral
8 are supposed to
...
be written
in
CWK,
general terms
^ + V, + ^3 + V* =
^3
v*
+ +^+ =4
+ + + =
^4
*'*
*;7
*'
v*
l*
/(',,
&,, &,,
k,
/' 4
= v,
= v,
6,
,
+ = v,
3
-f-
/-,
vt
304
+ 2*.
2k, -f
2/C- 2
=4
+ 2, = 4
+ 4^ =
/3
.
.
which
ma
=
=
=
=
.
'4
.
.
= *(+ 34 -*4 +
= i(+ 4 + 2/,= *(- 4 + 2/ +
= *(+ A-2/ 3
/,)
-f-
/,)
/,)
be written
() Write
sets
offour
the
measured angles
in order
second those of
of azimuth in two
of SCW, and the
WKS.
Adjust the angles of each set by one-fourth of the difference of this sum from 180 -f- excess of triangle, arranging
the adjusted angles in two columns, so that the first column ivill
(,9)
show
the angles
CWK.
to the
second column.
APPLICATION TO TRIANGULATION.
The
o".35,
Measured
angles.
SCW, CWK,
WKS
of (he quadrilateral
305
being o"i6,
may be ar-
306
still
more
rapidly as
that is, the corrections to the angles are numerically equal, but
are alternately -f- and
This plan has the additional advantage of not disturbing
the angle equations.
.
Returning
to
first
OSC =
33
53' 35"-40
OCW = 58
44' 39". 56
OWK=*$
54'
OA'S =27
38' 46". 88
58". 27
9.7463587
9.9318952
9.9326832
9.6665301
72
70
+
+
+
+
+
+
+
+
3I-3
7/1
7/3
7/5
2' 7
12.87/3
12.72/5
40.27/7
= 62
CWO = 25
SCO
WKO
A'SO
= 56
7/4
02' O5".i5
+
+
+
io".i7
7/ 8
03' 27". 82
18'
24'
17".
38
v,
z/g
APPLICATION TO TRIANGULATION.
307
the following
may be rigorously adjusted for the
side equation without disturbing the angle equation adjustment, which amounts to the same thing as the simultaneous
(a 2) Adjustment of a
artifice the quadrilateral
By
Quadrilateral.
+ <w' + "
",*','
writin
the
first
cally,
+ <w' + <*?*
-f
<***'
become
identi-
condition equation
("i
+ + a + "n
+
<* 9
(",
:,
<><'"
a.
(^5
a r )v -f
B )f'"'
(rt-,
rt-.j7''
^)''""
(",
/,'
with
(T'
The
40
= a mill.
308
Substitute
k\\(a, -f
in
a.,
from which
-f a -f a,
:>
a,
rf.
a^
(>,
can be found.
Hence the corrections are known.
/
Meas. Angles.
of our quadrilateral
is
APPLICATION TO TRIANGULATION.
For example, take the
first
(.'
n".~2)
Measured.
Adjusted.
Kingston,
Wolfe,
88
19'
14".
70
15". 78
55
5<y
20". io
21 ".18
Amherst,
35
41'
22". 69
23". 77
I7<)
5'V
57
oo'
oo .72
180
So"
309
o".
'-4'J
73 check.
23
3)3".
r.os
Fig.40
+ + +
'
'u
'
''
''
/,
Angle equations,
'
P,
'
<
',
*>
1
1:1
"l
'
v,
11
"l
'
=
=/
1,
'
)r>
We may adjust tor these equations in order, Hist the horizon equation, then the angle equations separately, as they
are not entangled, and next the sii.le equation.
A rigorous adjustment may, however, be carried out at
once with very little additional labor. Adjust first each
... be the values
angle equation by itself, and let (:,}, I;
denote the farther correcLet (l), (2),
that result.
tions to the measured angles in order arising from the local
and side equations, so that
1
..),
310
If
we
substitute
these values in
above equations we
the
+
+
(i)
(4)
/,
Calling k lt
equations in order,
,,
=
=
(2)
(1)
(3)=
now
Eliminate
//,
we have
+7
+7
*+7
)=o
-|- (i
=o
5)
(15).
,*>
,*,
(5)+ (6)=o
+ (14)
(13)
from which to
(2)4-
4)
(5)
(6)=
+ 77
(,
addition
By
+ ",X'i + 37
Hence
(2)=
(3)
=-!(,+
these values of
Substitute
equations, and
2
-Htf.-w.)^.-i
[aa]
we have
a,a t
,)*,
(i),
(2),
we
...
the
in
condition
Solving,
and thence
JX-,
=
=
[ii\k 9
[a]*, -f io/',
(i), (2),
3/'
3 /"
are known.
APPLICATION TO TRIANGULATION.
311
is
correlates
\aa\k,
(,
(a4
(a,
+ 5*.
+ #,+
+
+
*,
a>)
fourth,
for
.
/,
*.)
+3/7
*,
Substitute
=
+ ,)/ + (a + //+
+ //+...=/"
+ /
=0
+ 3/
//,...
equations
/'
=o
their
in the first
values
The
introduction of the
method
Geodiisie,
part
iii.
p.
93
Huge
in
pa isc hen
Nell in ZcitscJir.
Gradinessung, 1867, pp. 106 seq.
xii.
x.
I
vol.
vol.
Venncss.,
fiir
seq.
pp.
pp. 313 seq.
The process is not of any special advantage except in
;
directly.
shall
We
At
are
taken,
angles.
station
at
the
other
stations
the
locally
adjusted
312
Given Angles.
APPLICATION TO TRIANGULATION.
Corrections.
(!)
Adjusted Angles.
313
where
OA
the angles A,
still
evidently (Ex.
The
this
5,
p. 109)
ONSL
s. e.
of the side
OL
as derived from
(Fig. 19).
the base
NS
in
APPLICATION TO TRIANGULATION.
The chain
Station.
of triangles
OA S,
r
is
OLS.
315
occupied, and
i, 2,
3,
Let
be the station
and let A, B,
denote the
most probable values of the angles
which the directions of the different
tion,
most probable value of the angle between the zero of the limb of the
instrument and the direction of the
Fig.4l
in
in like
M = v'
arcs.
we have
APPLICATION TO TRIANGULATION.
317
at once.
They
+P: A +pi
x,
are
B+
[/vi/j
A', B
Ml,
...
A, ,..., and
X,
where
lt
J//,
',
.f,,
xv
(A), (B),
denote
their
ml'
ml"
= Ml' - Ml - A'
= Ml" - Ml - B'
ml'
ml"
-h
= J/ " - Ml - A'
= Ml" - Ml - B'
a
3l8
The
quantities ,f lf x.
being merely auxiliary quantieliminate them by substituting their values as found
from the first group of normal equations in the second
ties,
we
We
group.
have then
where
[],
rf
are
^]>
to
be looked on as mere
symbols.
In the cases that occur in ordinary work the computamay be still farther shortened. If we arrange the
tion
we have
group
A =A
A'=A"
/
//
= ./ suppose
=,/
"
and therefore
;//
in this
group.
APPLICATION TO TRIANGULATION.
319
Hence
if ///,
// rt ",
become
'
[<W]
= L/J-
//
->
where
/'
'/
-*T,*P
s
>'s
,/>"
at,
zero.
No. of
Group.
/"'"
[P\ [/'"'"I
The
written
coefficients of the
down
at sight.
now
be
320
is
If
By
addition, the
number
of arcs being
// s
i,
)==?
Hence
()
as
is
evident a priori.
148.
i.
= [;'"]
Checks of the
Normal
Equations.
The sum
is
number
number
cc
_]
APPLICATION TO TRIANGULATION.
/i" 2
+//j +/ i"/i"'
!
pi"
4-
[/]
-f-
r/,
[AJ
2.
>
[/"]
[/"]
// ==
[/>"]
-j- \
r/,
4 A'" ! 4 />2
[/]
2
.
/'-'
i,
^Jjj
ry,
'A'"
4 ...
...
^|
the
number
of observations
the
number
of arcs.
a
i
-^
[A]
and therefore
p'"] -}-...
4-
[/i]
Now,//
[/']
-$-
[A]
-f
The sum
[f/]
where ['/]
is
+ [*/] + L^] +
formed
in the
-[
same way
as
[^
For
since [/"w/"J
we
to
+ fi" + Pi'" +
pi
their values
[^],
[#<?],
321
+ [/'"''"] +
[A'.l
[A W J
-...==
^= o
322
149.
is
The
first
step
to find
We
No.
ol obs.
quan.
;/,
ft
n being the
number
To compute
\j>w\
L/
-\-u s
(tr a
of directions.
the value of
|
pvi<\.
147,
Eliminate
+ A V) +//"(#) +
Hence, as
in
x^,
re-
=[A<
Art. 100,
A
= [p,nm\
.r n
AJ
Iil-I
if possible, be found
from the original observations. It will in general be quite
different if found from the means forming the different
groups of arcs taken as single observations. (See Ex.,
Art. 62.)
APPLICATION TO TRIANGULATION.
the
323
\vv]
1^1
[',]'
\jnni\
-[aa\
150.
The general
ad-
justment
is
Art. 113.
Ex. At station Clark Mt.,
readings were
the
made with
Mi" -Mi,
Ai Claik Mt.
Spear.
Humpback
first
that
direction in earh
is,
MI"
Mi,
324
Assume
Take
the several
the
00
oo' oo".oo
Humpback,
24
09' 36". 90
Fork,
78
26'
9". 90
+ (A)
+ ()
the differences
/';'
of the angles
and
APPLICATION TO TRIANGULATION.
Next form
No. of
group.
the table
325
326
local directions
oo'
oo".ooo
Humpback,
24
09'
36". 975
Fork,
78
26'
io".O3O
Spear
150, is
found
to
be 25.7.
Therefore
the divisor
at the
at this station.
At Clark.
Most probable
directions.
oo'
oo".ooo
Humpback,
24
09'
36".975+(i)
Fork,
78
26'
io".o3o
Spear,
4- (2)
Weight Equations.
(l)= + O.I92I
(2)= +O.II30
I
|
1
|
[w]=25.7
+ O.II30
+ 0.2429
2
|
2
|
Divisor, 17
At Spear.
o"
oo'
Fork,
32"
08'
n".793+(3)
Clark,
54
06'
29". 197
Humpback,
oo".ooo
Weight Equations.
(3)= + 0.2061 jjj +0.0485 fjj
(4)= + 0.0485 |Tj +0.1879 IT?
\vv\
58.7
Divisor, 17
(4)
Proceed similarly
327
At Humpback.
oo".ooo
oo'
o"
Clark,
Spear,
101
44'
Fork,
332
58'
3".
123
(5)
n".i57+(&)
Weight Equations.
(5)= +0.1333
(6)= +0.0667
+0.0667
+0.1833
106.0
[rT']
f>
I
Divisor, 23
At Fork.
Clark,
oo'
oo".ooo
Spear,
79
35'
42". 479
Humpback,
98
41'
43". 926
+ (8)
(7)
Weight Equations.
(7)= +0.2970
(8)= +0.1394
Qj
+- o.i 394
8
I'
Divisor,
[<vz/]=47.5
+0.1879 |JJ
17
The
side equations are formed as already explained.
the triangles SFC (E
7". 386),
io".773),
(f.
HFC
angle equations
from
SI/C(f.
at
9". 789),
C), will
be (ound
to
CSHF
(pole
be
(2) -(3) + (4) +
-(i) + (2) -(6) +
(7)
-0.860
= 1.562
+ (4) + (5)
=0.494
+ 0.2187(5) - 2.0635(6) + 0.0193(7) + 0.1611(8) = 0.0424
(8)
(i)
2.609(3)
From
The
1.847(4)
this
finally
At Clark, Spear,
o"
Humpb-.ck, 24*
Foik,
At Spear, Humpback,
oo' oo".ooo
At Humpback, Clark,
844
Spear,
ioT 44
Fork,
09' 36".
78
26' io".47S
oo' oo".ooo
Fork,
32
Clark,
The m.
The m.
08'
At Fork,
n".799
s. e.
of
s. e.
o oo' oo'.ooo
Clark,
Spear,
79
Humpback,
98
41' 44".
is i'-77.
CSF =
o".46.
536
328
The form
for the
is
Survey
essentially
same as that just explained, so far as finding the values of the local corrections is concerned.
The method of
the
weighting employed
as
is
it
found
in Report, 1864,
152.
app.
14.
may
very
be derived from
X,
then
we have
= M,',
X,
= M,',
as approximate values of A, B,
\p"(M"
_
[/"]
[/'J
X^
,t'
X^
to
Then
the
_
[A]
which values substituted in the second group give as second
approximations to the values of A, B,
_ A'W: - M: +P:(M:- - M: .
and so on.
/)
Q+
T7T
W'-M';- Q +
APPLICATION TO TRIANGULATION.
329
We
then have
The
x and
corrections
made
as before.
Should the observation of the zero direction be wanting in any arc as, say, the third the quantity to be added
to each reading in this arc is given approximately by the
in the first set of normal equations.
value of
Thus, given
o,
to find
M: -
If A',
M,"
and
"
M,'"
- M '-
=A
approx.
first
two
arcs, then
330
Hence
M', M,",
M', M:,
A'
or
o.MSo,
M,"
A',
would be
M,'"
M,'"
M', M,'"
f
,
-M',
M/"
'"
M,'"
- M "+A + M'
f
M'
-M'
-M "-\-A'
Ex.
At Clark Mt.
APPLICATION TO TRIANGULATION.
331
though
The forms
the
cated.
of
work
method which
in the
work
as in
all
long arc; the more uniform light that may always be had
when the number of signals in use at one time is small, etc.,
we cannot but conclude that greater precision is to be
The
independently.
mutually balance. Even Andras,
the author of the most important contributions to the
method of directions since Bessel, and who used this
attained
errors are
method
43
in
more
the
likely to
triangulation of
Denmark, acknowledges
332
that " in place of observations of directions in arcs it is preferable to return to the old method of Gauss in measuring
*
angles."
As regards the cost, it must be acknowledged that for
an equal number of
out of account,
method of arcs has the advantage. Nowadays, however, when facilities exist for measuring angles by night as
the
measuring combinations
in
it
of the angles.
simple form that at once suggests itself
would be to close the horizon at each station that is, to
;
would each be
of the
LOA
in
order
local corrections
measure
all
adjustment
possible
for
this
sum
angles.:}:
Though
combination of measures
I
the
form of
a special
a little
risk
of
at
the
shall,
is
1878, p. 47.
t Sec C. S. Report 1880, App. No. 8. Experiments made at Sugar Loaf Mountain, Georgia,
have shown that an apparatus cheap and easily operated can be used that night observations are
a little more accurate than those by day, and <that the average time of observing in clear weather
can be more than doubled by observing at night.
;
App.
20.
APPLICATION TO TRIANGULATION.
154-
333
Let
(Angles).
be the
02
103
203
104
204
Take the
first
304
three as independent
Fig.
/13
43
denote the
are
(0
-A + B
-A
and the normal equations
(3)
It is
useful to notice as a
check that
A+JS+C=l + i + /
lt
tt
Sum
first
(4)
lt
in
tabular form as
sum
334
differences,
columns.
A,
APPLICATION TO TRIANGULATION.
335
i.
Given
at station
Oswego,
(Fig. 38),
/n
/13
/14
= 80
= 107
= I38
/as ==
4u=
Au=
in the triangulation of
29'
46". IO
ig'
03". 28
12'
49"-44
26
49'
l6".6l
57
43'
30
53'
oi".g6
42". 88
46". 10
would be
'
Lake Ontario
336
The angle
155. The General Adjustment (Angles).
and side equations are formed as explained in Arts. 117The connection between the local and general adjust139.
ment is through the weight equations. In our example of
4 stations, in which every angle between every two directions
is
of which the
first is (see
Eq.
2, p.
in three
groups,
333)
arms
of the
angles, the
developed.*
If X,
156. The Local Adjustment (Directions).
A, B, C denote the readings of the 4 directions i, 2, 3, 4
from O, then since A
X, C
X, B
correspond to the
* See
APPLICATION TO TRIANGULATION.
337
=
-X
+B
-/,. =
-JT
+ c-4 = *
-.* +
-/ =
-X+A
-/
*,.
I4
/?
+ C-/, = ^
-JB + C-l. = v,
-A
(0
*'
x- A- B- <:=-/-/-/
-,r^- 5- c= /-/-/
3
results
=
and therefore the unknowns cannot be found without some
further relation connecting them.
The reason of the indeterminate form is that directions are nothing but the angles
assumption
X+A+B+C=o
By adding
duce to
this to
(3)
/I 3
*A--
/_/-/
48=
/,,
4C
/,4
+ /,,-/,,
+ +
/,4
re-
'.,
C directly.
(4)
338
APPLICATION TO TRIANGULATION.
339
arms of the angles, the result is the same. One method or the other
may, therefore, be used, as is most convenient.
To avoid the use of large numbers certain approximate values may be
assumed for the directions, and the corrections to these approximate values
found. (Compare Art. Si.) Thus if (X), (A), (B), (C) denote the corrections
of the
to
assumed values
of
X, A, B,
C,
we may proceed
as follows
Duck,
Stony
Pt.,
oo'
oo*
80
29'
46"
+
+
107
19'
03"
+ (B)
138
12'
Tabular Form.
(A')
(A)
340
'
tri-
At Buchanan,
2-1
3-2
At Brule,
5-4
At Aminicon,
8-7
At Lester,
Measured Angle.
47
57'
36". 25
36". 21
97
26'
41". 29
4i"-32
49
29'
05". 15
5". 1 1
APPLICATION TO TRIANGULATION.
341
+o". 13
C1)
(7)
0".3I
(2)=
(3)=s+o".i8
On
=
=
o".i4
(i2)=:
+ o".3S
(10)
(5)=+o'.n
(ii)
(6)=
o".3i
(9)=+o".io
o'.os
(4)
(8)=r+o".2I
o".o6
o".24
the Breaking of a
would be very troublesome, not from any principle involved, but from its very umvieldiness.
Accordingly it is
necessary to break the work into sections and solve each
more or
By
method
the
much
This
itself.
less
is
a strong
argument
The
may
British
tions
115.
As an example we
Ordnance Survey.
to
be satisfied
in
p. 272.
in
the
342
The
out even
was divided into
carried
more systematically.
five
groups of
was
The adjustment
condition
The
IV.,
V.
For an interesting
see Comptes Rendus de
still
four
remaining were
r Association
Gcode'sique Internationale,
1877.
Circuit.
conditions
position.
in
Schwerin, 1882.
APPLICATION TO TRIANGULATION.
343
We
tion.
Adjustment
60.
of the
In
1-2
but
134
is
A A^ ...
lt
/?,,
}}
B^,
*,...
344
Let
z,
angles of the intervening triangles resulting from the adjustment for local and general conditions. Call the value of
the azimuth of 5-6
computed
in this
way
z' .
in
usual
-(Q + (Q-(Q + (Q = 4
where
*
(i)
equation
may
The unknowns
in
equations
I,
are subject to
the re-
lation
'
-
Call
,,
k^ k^
= a rain.
=
k, + k = (Q
*.
\
*
(Q
k.
- k = (Q
+ ^ = (Q
1873
Coast Sur-
APPLICATION TO TRIANGULATION.
345
=o
k=o
k
3^
3^3
-f
*=
2n
~
and therefore the corrections
Hence
the azimuth
line of
34^
The
sides 1-2, 14-15 are the lines of junction with the primary sys-
These lines are assumed to remain unchanged in azimuth in the adjustment, and the secondary system is to be made to conform to them.
The main chain of triangles joining the two primary lines is indicated in
The system
the figure by heavy lines. The number of these triangles is n.
has been adjusted for local and geometrical conditions, and the resulting
angles of these triangles are as follows :
tem.
Angle.
APPLICATION TO TRIANGULATION.
347
az. of
computed
14-15
az.
of do.
2".
93
Hence
triangle
o". 13
(/!,)
(/>',)
=-o".
(Q = -f
= -\-o". 13
=+
W
(Q =
".26
o*.is
o".26
and so on.
161.
Adjustment
Arts. 168-170.
fully explained
Using the last form given in Art. 170,
base-line equation
This
in
is
\dA
(A)-dB (B)-\
we may
write the
=l
and Longitude.
The
in
45
348
The discrepancy in position, as shown by the differences between observed and computed latitudes and longitudes,
This discrepancy, being small, may be
eliminated closely enough by distributing it proportionally
from one end of the chain of triangles to the other, according to Bowditch's rule as given in Ex. 5, Art. no that is,
alone remains.
several stations.
eccentric,
is
will be consistent.
CHAPTER
VII.
compensation bars
On
and
the
metallic-thermometer
appa-
the English
is
surveys.
* See
Ibanez, Zfitschr. fiir Instrumentenkiinrif, 1881.
Also Art.
166.
350
164.
clearness
it
will be
For
made
at various
ture
roneous.
351
compare
results at the
end points
only, \ve
may compare
way a better idea of the precision of the work is
obtained, as we have 6 or 8 short bases to deal with instead
In this
We
A' A'
m.
s. e.
>
F,
we have
= most prob.
first
section
First section,
V,
M{
?>/
V,-M{'=,v;'
Second
/ P" A">
/
respectively, required the
of the most probable value of the base.
Let
then
section,
F,
J//
- M,"
V,
r>/
= v,"
wt.
/>,'
wt.
A"
wt.
//
wt.
A"
and so on.
Now,
either of
352
(a) In
the
place, that the precision of the measureis the same throughout the different
first
whence
line
F,,
V=
V, -f
are
-f
unity
by
is,
of a single
of an observation of weight
of a bar
is given
/j.
measurement
No. of obs.
Now,
unit of
number
For
of bars measured.
since
n VM.
length
length
ft
s. e.
of a measurement of a single
bars
of the measurement "of a length of
is
Hence
the m.
=^
is
s.
the
e.
weight of a measurement of
of a
unity.
Vvi
vv~\
(,-
M\
i)\Jf
in
this
formula
in a
w,
For
if
M M
lt
the
t,
first
.
.,
353
measurement
a bar
for the
and
for
the
of the
mean
of the
the
number
354
Hence
the m.
s. e.
measurement
is
only,
y_ "" .094'
6
87
and the m.
s.
0.093
e.
283/.007
92
-+-.cos'
60
L\
+ --757'
131 /
of error are
1.
2.
The m.
s. e.
Remembering
m.
--
s.
e.
0.386,
0.391,
0.335
f r
of base
= Kogs^
=
.327'^+ (.386
we
.391)'-
have, finally,
+ (.313 + .335)-
L
1.07
mean
for the
m.
s.
e.
J&wL
v
,,=
since
Sincea
A =ff=.
_
n
/
-,/
[AJ<X
i)
,(,
i)
V denotes
the whole
line,
so that
r=* ,+ * r.+
r
The number
+Vn
s.
355
e.)
of a single
measurement of the
line
for
and therefore
and the
(in. s. e.)
46
of a single
measurement of a bar
is i
FTT
356
Ex. The Chicago Base, measured in 1877 with the Repsold metallicthermometer apparatus belonging to the U. S. Engineers. The base was
divided into 8 sections and was measured twice.
Section.
was
"
that
"
The
We give
net
is
shown
in the figure.
3077459
2216.397
Triangulation.
Fi 9- 47
3077-462
2216.399
2766.604
2/23.425
2723.422
3879.000
3879.002
14662.885
14662.889
Flasis.
Speyer, i8?z.
Astron.
A'ac/ir.,
No.
1462.
358
"
two operations
the
neglected,
measured n times, and the arithmetic mean of the n measurements taken as base from which a line equal to nM is
derived by triangulation, the precision of this line is the
if it had been measured once directly.*
if
the m. s. e. of a length M, then f* M V7i is the
M
m. s. e. of a length nM. Also, the m. s. e. of the arithmetic
same as
For
[JL
mean of n measurements
from
m.
nM
s. e. is
.-V
of the line
= HM Vn
is
is
7^,
Vn
and when
derived trigonometrically
Jl
We
its
may
amount
tri-
of influence of
Assuming
we have
(Ex.
9,
Ha
where aH
is
the
/V
computed
triangles intervening,
-J-
f // sin
\"
Vn
and
//
the
m.
s. e.
number
of a
of
measured
angle.
=
*
fj-b
+-1600
Gradmessung
nfjf
millimetres.
in Ostpreussen, p. 36.
10000,
359
But little is gained in precision by repeating the measurement of the base many times. We have seen in Art.
164 that the main sources of error to be leared arise from
the comparisons with the standards and are independent of
the measurement of the line.
Accordingly, though bases
have
in
times, it
In this
is
error
checked and a
is
sufficiently
practice
wanted as triangle
tries will not
*
The
p. e. of
sides, as the
allow of
it.
is
part
of the Grossenhain
Base
5865800
1872),
Base
Saxony,
(1877),
U.
part
of the Atlanta
part
S.,
of the
Base (1872-1873), U.
Sanduskv Base
(1878),
U.
part
S.,
Harlem
7.0 kil.
kil.
Halland
60
(1863),
Sweden,
7.3 kil.
Atlanta (1872-1873),
Udine (1874), Italy, 3.2 kil. ; Chicago (1877), V.
(1868-1869!, Holland,
9.3 kil.
kil.
Chicago
1148600
of the
part.
S.,
1052200
Oran
(1880),
Germany,
kil.
5.2 kil.
Meppen
(1883),
S.,
S., 7.5
Germany,
360
166.
Measuring a base
The
line
is
is
the measurement, as
that of melting ice.
its
temperature
is
at
any
under
and
Connection of a Base with the Main TriWe shall now consider the best method of
aiigulatioii.
167.
S. A.,
No.
the Rep-
361
is not to be decided
by least-squares'
methods alone.
There are two points to be considered as little loss of
precision as possible, combined with economy of work in the
measurement and reduction of the intervening triangulation.
To secure the first we must use only well shaped
geodetic questions,
the bases.
should be occupied.
The
nection
solution
may
is
a tentative one.
s.
e.
found
sults will be
shall
now make
in
use
results
we
side to be from
15
These
of.
measurements,
that
form
AB
the base
is
numbers
16, 3, 7.
Hence, so
is
with the
first
form.
that
it
362
can be found by passing through two sets of similar triangles from the base, is the normal form, and that the forms
used in practice should approximate to it as closely as the
nature of the country will allow. As examples take the
connections of the Coast Survey Atlanta Base (1872-1873),
the Lake Survey Chicago Base (1877), and the Prussian
Gottingen Base (1880).
Fig.50
Fig.49
Fig.48
ATLANTA
QOTTINQEN
when more
Adjustment
The
be found.
How
shall
made
in
363
such a way that no discrepancy would show itself in passIt was only necesing from one measured base to another.
to
introduce
sary
equations connecting the bases in the
form of an ordinary side equation, thus (Fig. 12),
where
are
a, b
two
sin
sin B^ sin B^
bases,
and
sin
A B^
lt
continuation.
There
adjustment.
equation to
be
written
<*
where
+ (*)
+
sin \A,
sin
(b]
a, b,
t ,
Z?,,
+ (Aft
sin
\A
+(A
t )\
(a),
47
(/;),
d.(a)
4,
364
where
usual.
with
\
where
m.
= a mn
the
are the m. s. e. of the bases, and //,, // 2
/Jtb
of the three angles of each of the triangles in the
/^a ,
s. e.
chain.
The
solution
may be
method
of cor-
relates.
If
maybe
illustrated
BC = 6742.420
AC = 6602.386
Sandusky Base,
ft.
o.oio
o.oio
From
07' 51". 35
o".2o
ABC=
06' 26". 74
o".20
ACB =
177
45' 41". 91
o".2o
Required the most probable values of the sides and angles of the triangle
from a second adjustment into which the two measured sides enter.
365
If (a), (/>), (A), (B), (C) arc the corrections to the above
quantities in order,
the condition equations are
6742.420
6602.386
(A)
(a)
_
~
+
+
(B)
sin
reducing the
o.644(<z)
that
(a)*
+ W_ +
(.01)-
(.01)*
+
4-
(A)\
(B)\
(JBT-
(.20)-
i.obi(B)
(C)
0.044
(^ -
(.20)*
is,
2
4oc<rt)
The
07' 51". 35
06' 26". 74
i.oS()(A)
W+
form,
+ 0.658(6)
with
=o
sin
(/>)
or,
(C)
40o(/')
+ (A)- + (B)- +
(C)-
=a
min.
= + 0.00003
= 0.00003
(a)
(t>)
=
=
(C) =
o".O2
(^)
(^)
o".O2
o".oo
(2)
The two
Approximate Solutions
of most importance
when
are
justment.
(a)
When
to be used
method
zero.
The formulas
the angles alone are changed.
this case follow from those of the rigorous
in
just given
Thus,
if all
(a]
= o,
(//)
becomes
with
are as before.
o,
366
Hence
is
(Q=(A,}=
(V(20Y
we
dB '}k
(2)
dB "}k
= - OY- V)*
.
whence, by substituting
Hence
It is
still
evident
A=B.
we may
take
= (Q=.
=o
(3)
whence
.^_
'^B_ k
2
where
(5)
367
Ex. i. To find the changes in the angles resulting from the equation
connecting the lengths of the lines 1-2 and 14-15 in the triangulation of Long
Island Sound (Fig. 46).
The excess of the log. of the observed v.ilue of 14-15 over the value-
computed from
decimal place.
If the lines themselves receive no correction the condition equation,
expressed in the linear form, is (see table, p. 346)
0.27(^1)
o.99(j9i)
o.6o(B-i)
i.6i(/l a )
4.00
Now,
[(8 A
+ s By>]
51.48
and therefore
(A
= -(B = 0.27 +
l
o.qq
^X 4-00=0". 10
Ex.
2.
To
two
bases.
Having found the adjusted angle?, we may find the weight of the log. of
any side of continuation as the m th in a chain of triangles joining two bases,
all of the angles being of the same weight.
For simplicity in writing take in =2, n
3.
are
3)
/',
whose weight
- 8 B "(B,)
at
UF = f [8 A
+ 8 A "(A 9 )
15, Art.
a
>
summation of
- 8 B "(B*)
The
in.
8 A '"* + 8 A
+ 8 A d B + 8 B*]
[<>A*
the
- 8 B "\B,)
be found,
is to
+ 8 A '"(A 3 )
between the
"
t>A
result
is
"'SB"
~ + 8,B
8 B + 6 B -]
t
limits indicated.
368
As an example
of a different
method of treating
this
the
base,
is
we recede from
it,
obtain a proportional
amount
of the
two
lines as
we
ap-
proach them, and finally, reaching one or the other, the effect
of that base from which we set out is lost.
Each distance
will have assigned to it its most probable value
we shall
have no discrepancy whatever in the geometrical figure of
the triangulation, and the resulting sides and angles will
have nearly the same probability as those derived from a
;
is
that as the
ancy
is
This
is
the
21.
Pv
b.
,,
... be
369
their weights.
the most probable length of a
.
If s is
/,, / 2
common
side
com-
y
/
weight
=r
'v
M'
value of log
s,
is
pjb*
is
/.A*
X,
log A
<i
With
p}>?,
s is
may
We
is the main one, and that the error arising from the
measurement of the line itself is relatively small. Hence
bars
we may
take the m.
As
s.
e.
we have now
is
evident that
370
Thus
in the
to,
CHAPTER
VIII.
APPLICATION TO LEVELLING.
As only
differences of height,
mean
To
level of the
be able to refer to
it
at
mainly
in the
in
the
latter case.
For descriptions
instruments employed see AYpar Hirsch et Plnntamour, Geneve et Bale, 1867, seq. Report Chief of Engineers
U. S. A., 1880, App. GO; Report U. S. Coast Survey, 18/9
App. 15, 1880 App. II Professional Papers Corps of Engi171.
of
neers
U.
4s
S. A.,
No.
24,
Chap,
xxii
372
The method
as follows:
Two
run simultaneously, with the rods usually at different distances from the instrument and to prevent the gradual
accumulation of error supposed to be due to running constantly in one direction, alternate sections are run in opposite
;
Each station of the instrument, therefore, contwo backsights and two foresights. In stations i, 3,
directions.
tains
Rod
13,
from
"
May
"
14,
"
"
"
"
9,
10,
19 to
20 "
24
24
Aug.
18,
0.51
15,
0.46
6,
0.37
6,
0.43
"
"
Sept.
"
mm. per
metre.
"
"
"
"
is
necessary.
It is a curious fact in spirit levelling, but one abundantly
verified, that a much greater discrepancy is to be expected
in
over
vol. v.
Berlin, 1883.
in a line
APPLICATION TO LEVELLING.
373
When
rod No.
Distance.
374
Again, as both rod and bubble should be read simultaneously at each foresight and at each backsight, and as
this is physically impossible for one observer, the effect of
the unequal heating of the instrument by the sun, even
when shaded by an umbrella, has a tendency to cause a
change in the position of the bubble in the interval between
This error is cumulathe readings of the rod and bubble.
minimum
tive, and its reduction to a
depends upon the skill
of the observer.
no case exceed
mm.
On
\^2D
mm.
is
allowed.
The
precision
of a line of levels will be given by the m. s. e. of a single
levelling of a unit of distance, which we shall take to be one
The problem is quite analogous to that already
kilometre.
172.
if //
is
the m.
s.
e.
of the
APPLICATION TO LKVELUNO.
unit of distance (one kilometre), the m.
D kilometres
s. e.
37$
of a distance of
would be
in
kilometres
would
m.
would therefore be as T
O'J
attendant
on estimating these sources
uncertainty
of error is so great, and the influence of the distance so
The
largely exceeds the influence of the others, that it is sufficient to adopt the rule first given of weighting as the in-
verse distance.
It
work
is to be looked for
short sights are taken.
Even with a
first-rate telescope sights should not be taken to exceed 100
if
metres.*
As
the rod
is
more
easily read
by the observer
at
appear.
On the precise levelling of the Coast Survey, " where the slope of the ground is steep the
distances may be taken as great as possible, and on comparatively level ground they may range
from 50 to 150 metres, according to the condition of the weather and atmosphere."
*
The U. S. Engineers follow the rule, " The lengths of sight will depend on the condition of
the atmosphere, but the rod should always be near enough to be seen distinctly. It will be seldom
that lengths of sight greater than 150 metres can be taken."
In the Prussian
Only
in special cases
shorter sight
is
example,
frequently taken."
in
3/6
then
if z//,
v",
v,',
vt ",
...;...
in Art. 164
the hypothesis that the precision for each unit of
distance (one kilometre) is the same throughout the different
On
(i)
sections,
(n,
i)
value of
fJL
When the number of levellings is two, and the observed differences of height at the several bench marks
Dn kilometres apart give discrepancies of
Z>,, Z> 2
dlt d^
dn respectively, then the above formulas reduce
.
to
and
The m.
s. e.
respectively
and
and
for the
m.
s.
e.
of the
mean
of the
two measurements
APPLICATION TO LEVELLING.
377
B.M.
378
Those arising from non-agreement of repeated measurements of differences of height between successive bench
(a)
marks.
if s
the
number
of
number
satisfied
among
j-f
is
i.
1
74. Approximate Methods of Adjustment.
of height between successive bench marks
The differences
may be found
APPLICATION TO LEVELLING.
379
for
quantities,
(a)
The
marks
will
non-satisfaction
of
in
of Art. 109.
versa.
racy
is
(d)
vey in
in
reached.
380
common
as a
of equations with 91
unknown
system
quantities."*
and the
zero,
A 1^=42.65101
Z>i
AM
Z> 2
AG
Z> 3
10530
Z> 4
=44.2
3.79892
Z> 5
=58.56223
WM =
Fig. 51
12.
MG =
=37.8
= 35.8
= 22.6
= 54. 74663
kil.
"
"
'
"
27.9
First Solution.
z/ 2
z>a,
v\ -=v\
=
=
=
z5 =
v<2
z/ 3
z/ 3
z> 4
five
The points
7' 5 .
We
Vy
z> 4
IV,
measured
M, G
v\,
are
z>i,
v3
weight 26.5
"
27.9
44-2
"
+ vi 0.00968
Vi +
+ 0.01668
vi
"
z>s
22.6
35.8
1000
the weights being
The solution
computed from
is
finished as in
Second Solution.
From
first
A JVM,
AMG
we have
condition equations
v\
V1-\-V^=.
Vi
va
Vi,
0.00968
O.OI668
with
J>\
The
solution
is
* Abstract
Introduction, p.
vi.
the
APPLICATION TO LEVELLING.
ThirJ Solution.
should be zero, we
may
We
AM=
54-74663
MG =
3-79892
AC =
58.56223
D*
>
0.01668
Correction to
AM =.
35-3
.'./.,
27.9
22.6
pb
/3
86.3
0.01668
= -{-0.00692
06.3
Weight
of adjusted
AM = 27.9
= 47-7
Hence
-f-
35.8
44.2
AMW
*
^^=42.65101
I2.I053O
=47-7
=26.5
=22.6
The
circuit
Fourth Solution.
381
and so on.
=
=
27.9
35.8
= 44.2
382
Trigonometrical Levelling.
In extensive surveys in which a primary triangulacarried on, the heights of the stations occupied may
175.
tion
is
When
is
but slight.
a country
more
de-
See, for
of
of
the
two
the relative accuracy
methods,
comparisons
Report of U. S. Coast Snrvty, 1876, App. i6and 17 Report of
G. T. Survey of India, vol. ii.
Report of New York State
;
of inaccurate
work
in
trigonometrical
is one of those
atmospheric
erratic
in
its
character that
disturbing causes which is so
no method has yet been devised for determining it that is
levelling
refraction.
is
Hence
very satisfactory.
metrical levelling
minimum
may
plan adopted
in
trigono-
is
refraction.
crepant results
the
This
Without
be looked
this
for.
For example,
in
India,
"
vertical
APPLICATION TO LEVELLING.
383
taneously.
A^ A
In
the
figure,
if
two
But on account
Z^A^A^.
refractive
power
of the
of the atmos-
A^A^
and the
Thus TA A,
t
Mem.
is
384
s,,
lt
D=the
C
'
: ,
by D.
Now, assuming the paths of the rays of light from one station to the other to be arcs of circles, which is approximately
the case when the lines are of moderate length, we have
C
_ ^~path
t
of ray
:
7>
c
r_
path of ray
~~R~~
where
C.
where k^
k^
of the earth.
= iV
C,
= i*,C
(0
be called refraction
may
factors.
Observations are frequently made so as to be simultaneous at A A^ and the line of sight may then be assumed to be the same arc of a circle for both directions.
t ,
In this case
if
we put
IC
K
^
r .\-r
TSa
(si
is
The
triangle
C= R sin
:
in
Tl
question.
where
is
the
APPLICATION TO LEVELLING.
Hence, attending to equations
I,
385
Art. 177,
we may
write
+ ^ = 2|r- -^-(^+ ~ -
*1
80)
(2)
|
This equation shows that a single line will not give the
refraction factors, and we must, therefore, have a net of
lines with the zenith angles read at the ends of each line.
If,
have
ABCD
we
shall
Fi g .53
unknowns
reduce this num-
we may,
order to
ber, assume, if the observations at station A over the lines AB, AC, AD are
nearly simultaneous, that the k for each
in
of these lines
is
that
is,
the same.
most
tions.
not always be possible to get one k for each stahave been sighted over,
in
number of equations over unknowns
a sufficient excess
It
may
tion, but
of least squares.
these equations,
.cr,
given by
that
is, its
",
386
179.
When
Coefficient of Refraction.
= = k suppose
,
at the
moment
of observation
,.
A value of k can thus be found for each line sighted over
from both ends simultaneously. To get an average value
of k for the whole system the weighted mean of these
The same method of
separate values must be taken.
used
as
in
be
the preceding. Bessel
assigning weights may
from
that
errors
arising
irregularities in k are of
argues
much more importance than the errors
and proposes the empirical formula*
In this he is followed by
for assigning relative weights.
the Coast Survey in their determination of the coefficient
of refraction from observations made in Northern Georgia
The mean
80.
There
When
used.
(a)
When
Let
known
//,
//
=
*
Gradinessung in Ostfreusscn,
p. ig6.
APPLICATION TO LEVELLING.
Krom
the triangle
CA,A
H.-ffi
Substituting for
CA.
A A
t ,
387
(Fig. 52)
CA,
tanKAt-AJ
zv
(}
kv
and reducing,
When
(b)
and
which
is
When
(2)
used separately.
common
In the
only,
we
have, since
= Z?
which
is
cot
^+
Z)
in (i)
and reducing,
-=& cof ^
When
two values
(4)
we have
388
2,
7z
= /Man i<> - O
= Z? cot + -,)!
(i)
#,
(i
/jt
(2)
These
182.
Levels.
APPLICATION TO LEVELLING.
389
The same
observation equations
(
^=2
R sin
i"
The condition
to
(*
y:
180
#,
be satisfied
|-
amon^
o
/,,
known quan.
the differences
of
stations,
we should
Thus
find the original height.
the triangle in order of azimuth, if//,,
ing round
note the differences of height of the stations, and
(// 3 ) the most probable corrections to these values,
proceed//
// 3
(//,),
de(// 2 ),
we must
have
that
we must have
is,
ak^
where
-j-
bk^
-j-
ck^
/4
c, / are constants.
four equations may be solved by the method of correlates, and the differences of height may next be computed
from equations i, Art. 181, and will be found consistent.
a, b,
The
the circuit, instead of being a simple one, has diagonals, then, as in Art. 173, if / is the number of lines read
over, and ^ the number of stations in the circuit, the number of conditions to be satisfied among the differences of
If
height
is
l-s+i
390
Stations occupied.
APPLICATION TO LEVELLING.
We
Observation equations.
(i)
'
--~
BD
2Ji
B=
where
J?
sm
The weights
Ij
are
computed according
enough to take i,
the best we can do with our
will be close
That
391
is
equations
1
a
2
(2)
+
+
+
+
+
=0.249
3
4
=0.317
as the
- V
formula
D (Art.
179).
number
of days of observation.
have then the observation
We
data.
= 0.267
= 0.308
= 0.297
to the
weight
"
"
"
1.04
1.33
1.35
0.80
"
0.85
Condition equations.
.-.
=5
=4
=5
sums
of
two
o=
o=
The
10.92 ki
1.52
1.84
14. 47
4 1.79
+40.17
52.71
54.64/^4
[] = a minimum,
gives
1=0. 1 18
= 0.107
whence
= 0.149
= 0.171
m
1
to 3
3 to 4
4 to
2.30
4 to 3
The precision
+
+
191.17
188.88
check sum
o.oo
2.30
found exactly as
}-
195.94
3 to 2
2 to 4
198.24
may be
392
183.
On
the refraction
method
factors,
is
it
lowed.
In the ordinary method of observation, where the observed zenith angles are simultaneous at every two stations,
//
H = D tan
= H,
| (z,
- z,}
and the differences of height may be computed at once without any reference to the coefficient of refraction. These
differences of height, considered as observed quantities,
may
in
the net, as in
The weights
P to
When
the zenith angles are not simultaneous, after finding the refraction factors, as in Art. 182, and computing
the differences of height, we should find the weights of
these differences of height from
It
would seem
safe to
Now,
Z> sin
as
Hence
first
assume pz
for distances
term
is
2",
p. k
0.02.
1"/** <
>
| ~R* P*
4 miles.
APPLICATION TO LEVELLING.
the second term.
We
393
amount
Ex.
i.
St.
stations were determined by trigonoThe following were the results unadjusted with their
metrical levelling.
Required the values adjusted for closure of circuits.
weights.
37) the heights of the trigonometrical
Diff. of height,
DifT. of height.
Wt.
394
among
where
/ is
Ex.
show
the
4.
If
number
and
the
number
= Di
cot 22
D\
cot
z\
"2.R
where D\, Z> 2 are the distances from the station occupied.
Also show that when D\
Z> 2 the precision of the height h is the same as
if the observations had been made simultaneously at the two extreme stations
themselves.
CHAPTER
IX.
Line Measures.
Let
AB
be
is
no
measure
line
length, so that
the distance
is
Fi
55
*
I
1
between o and n
to rind
is
known.
in
common
(a)
use.
Two
possible equal
the distances
o,
1,2;
distance
51
is
396
tween the micrometer zeros, which differs from the distance of the true positions of the marks by a fixed but un-
known amount.
Let
.*, x^
marks at
tion
if
xn denote
o, i,
M M be the readings
,
stant distance
at o, I and c be the unknown conbetween the micrometer zeros, and d the dis-
J=
*i
(0
4,
where
fol
= M,- M
and y
= c-d
<>
'
\
(2)
tional
observations involving other combinations of thein order to apply the method of least squares.
unknowns,
1}
.,
we have from
compared
first
397
space
'
(3)
Zi
^l
t*
But
=
(4)
Hence
V
3^0
l)
(5)
Also
-r,
^,,
.*.,
.7,
-|-
->
Ex, In order
marks needed
to
results obtained.*
m
Interval,
With
in
= o.o to 0.2
0.4 to 0.6
o.o to 0.2
in
2.6
o.i
0.6 to o.S
0.7
o.
0.8 tO
4-
2.2
0.1
= o.o to 0.2
I.O = 0.0 tO 0.2
2.1
o.i
=
=
The
+
+
+
zi
tit
0.2 to 0.4
JO*
1
.
65
t,
where
/ is
Sa
2.6
34=0.7
1.5,
O.6
and
xi
= 1.5
JT 2
r=O.
German
standard, to be
398
or in tabular form,
Graduation
But x and
;r 4
4*1
JT3
*3
Xi
+ 4X3
^3
Xi
Xl
JTl
JTl
}'3
I'*
}'3
la
/OS
/()
A.
4,1'l
JT3
yi
4X3
+
+
are known.
399
3^3
+ 2J 3
JT 3
=
=
/I 2
+ A3 +
/l 4
+ /8
AS +
/S
/2
/2 3
A) 2
/3
/u 3
/3
/3 4
(?)
may be found.
The Precision of the Corrections x xv xy The m. s. e. of
an observation of the unit of weight is found from the usual
We have
formula.
rections
-6
the
Chapter IV.
The weight of
Hence
\-.
x^
and of x3
s.
e.
of
x x^
method
the
example
the m.
Ex. Solve by
this
: ,
V,
is
ar| d
the weight of
is
ti
x^ are known.
in Art. 187.
(b) The work of reduction is much facilitated by employing an auxiliary scale, CD, divided into spaces approximately equal to those of AB, and whose values have already
been found by comparison with some standard.
as
If,
we suppose
all
400
.f u ,
;t-
as
unknowns.
Taking-
o, ,r 4
o,
the normal
equations are
-
from which
X?
Xa
JTl
+ 4^2
*3
-l"l
#2
4-* 3
=
=
=
+A +A +A
+
+
/OS
As
+
/o
/I 2
/2 3
/a 4
/3 4
/O 2
<'a
(8)
;r,,
may be found.
The reduction may, however, be made more
;tr
2,
;tr
easily
by
may be
written
X +
X
JTo
Xi
Xa
X\ "=
x<i
xa
xt
Xi+ 4X3
Xa
X.i
+ 4-^3
Xl
Xi
4-*o
4x1
fl ~~
-^2
=
=
=
Adding
'03
+ /o +
+
+ +
+
'13
/O 4
/o
A> 4
/o
+A2 +
/o 2
/i 3
l\ 4
/I 2
/2 3
/2 4
'2 3
'3 4
/2 4
/3 4
(9)
+ Xi + Xy + X + X =O
3
=+
=
5^2 =
/O
5*1
lo
The whole
tabular form.
solution
/O 2
'03
5^8
5.3:4
to each of the
this relation
5^0
+ /O + /O + /O
+A +A +A
A + /2 + /2
2
As
'23+
=-/ -A 4
may be
The sums
(lO)
/2 4
normal
(J J)
'34
/3
conveniently
of the horizontal
arranged
rows
are
in
first
The
vertical
401
402
it
could be
filled
by a certain volume
scale.
as nearly as
the positions o i, I 2, 2
in the positions 02, i
positions
As
it
03,1
is
3, 3 4.
3,
2 4,
4.*
quired
as possible
86.
* This
method
of
is
known
as
Neumann's.
403
little
column
way than
in this
in
the other.
It is
length were so obtained in this cnse.
rather difficult to break off short columns 5 to 15 in length
in the manner first described, the weight of mercury in the
short column not giving momentum enough to move it
away from the rest of the column readily. A little patience
To be able to read the
is all that is necessary, however.
shorter columns at 32 a column 10 to 50 long, depending
on the temperature at the time, must be broken off and put
about 160
in
way.
187.
52
Let x
-i',,
.r 3 , ,r 3 ,
x^
404
tions at
mately,
marks.
Then, approxieach interval a relation of the form
we have
column
for
diff.
of read ings
ofcal. corr.
-(- diff.
will
correspond
in
form to equations
Art. 184.
The
is,
however,
rather complicated, and though in comparisons of standards of length it may be allowable, from the precision with
which measurements can be made, to spend the labor de-
manded by
labor
Thus
in
if
denote the corrections to the 4 inte/vals
4
,,
our example, then
,
- -Y
f -- 'Yt 2
"Z
From
ing
"~
"4
*!
-Y -*4
we
have,
in pairs,
_ r-
""4
/
34
*a
*0
*i 3
- s
-"
'
-i-S
/
*
( T
l ^
\ J
by subtract
405
/
/33
/)2 ... as
Hence, considering /ia
independently
observed quantities, we have the normal equations
,,
-1
Z\
22
;3
3C'l
23
2l
23
2,
2a
2j
(/i 3
/oa)
/o ,)
/, a )
+ (A + (A -
/o 3)
24
/, 3)
a)
/a a)
(As-
/o a)
3)
(A
-4
323
/o i)
= - (A + (A - A + (A
2 = - (/a
- (A - A - (A - A
3=4 =
3)
(3)
- (A - /o 3)
4
Adding
this relation
4=1
423
4c3
434
2a
SS
to each
+ Si = O
of the normal equations,
unknowns.
= (A - /
= - (A - /
= - (A - A
(4)
- / ,) + (/,,- /
+
+ (A - A + (A - A
- A 3) - (A - A.
a) + (A
_ ,) _ (/, _ /
= _(/, 4 _/,,)_
The computation
,)
(/, 3
of the
2)
(/., 4
/,
3)
3)
3)
-1
we
Thus
facili-
406
2,
we have
for the
m.
s. e.
of a column length
6-3
number
knowns 3.
the
Now,
since the
m.
m.
s. e.
s.
e.
result by
V2
thus
2(6
- 3)
of
them
Ex. The following were the observed values of the lengths of the 45, 90,
and 135 columns of thermometer Green 4470, made to determine the calibration corrections at the 77, 122, and 167" points
:
45
col.
.,
Arranging
in
tabular form,
407
CHAPTER
X.
variable.
ent problem.
From
observation
from
made with
a staff
self- registering
409
From
M, =/ (x, a,
M, =f (x, /i,
t
where
and
J/,,
a, b,
//,
&,c,
k,
I,
k,
x the
variable,
In order to apply the method of least squares to determine the constants, the functions must first be reduced
to the linear form, and the number of quantities to be determined must be made less in number, arbitrarily it may be,
than the number of observations. The first point is to
determine as nearly as we can the general form of the
functions connecting the observed quantities and the variable.
For
this
considerations
make
ing to
known
The
we
in
are seeking. This method of eliminating irregularity by drawing a mean curve to represent the value
function
4IO
common form
for
parallel
"
It
deducing an algebraic expression for it.
manifest that some further combination is necessary
to allow of
is
in
Taylor's theorem,
/(*)=/{*' + (*-*')|
=/(*')
where
b, c,
*
Physics
are
constants to be determined.
Mississippi River.
Washington,
1876.
41
The question
down
in
number
of
terms of a
VM-\-b(t-f}
where
the
first,
so that
to represent
it,
V
A
we must
thus
M+
second guide
is
b(t
t'}
c(t
tj -f
sum
is
we may
still
unsatisfactory
in
func-
412
when
we cannot say
all
is
192.
It is
may
derive formulas widely different to represent like phenomeiia, and that each formula may satisfy the special set of
As an example
of this
we may
cite the
formulas proposed *
different temperatures
vapor
/.
Young proposed
(a -f
/;/)"'
Roche,
P=
where x
/ -\-
aa
mx
20.
P=
-f-
Mousson, Fhysik,
vol.
log
bo*
ii.
cjf
Zurich, 1880.
at
Of
first
three,
two observations
five.
If
involves
two
The
first
five.
to
determine
group
413
constants, the
curve, thereit,
the second
of observations were
plotted with temperatures as abscissas and pressures as ordinates, then, since the last lormula represents a curve passing
through five of these points, it is evident that if the obser-
vations are
near
all
good
fixed by only
this
two or three
points.
It
should, therefore, be
chosen.
Ex. i. The following are the results of the observations m;ide for velocity
of current of the Mississippi River by Humphreys and Abbot at Carrollton and
Baton Rouge in 1851. Each is the mean of 222 observations and is given at
proportional depths, the whole depth being represented by
below surface.
unit}*.
414
Hence
=
3.2299 = a + o.ib + o.oic
3.2532 = + .26+ .04<r
-
3.1950
To
Take
tract
the
mean and
sub-
we have
=
+ 0.0537 =
+
0.0188
+ 0.0770=
+ 0.0849 =
0.0754
0.45*5
0.2851:
-35^
0-275<:
0.25^
0.245^
o.
15$
o. 1951:
o.
o.os/>
i25c
= + o.os/; O.O35C
+ 0.0045 = + o. 15^ + 0.075C
0.0496 = +O.25/; + 0.205^
0.1168 = + 0.35^4- 0.355^:
0.2003 = + 0.45^ + 0.5251:
+
Hence
0.0520
0.2031
= 0.8250^ +-0.74251:
0.2232 =o.7425/>
and
= 0.4424
.'.
The
tions,
and give
c=
3.1952
o.
0.7652
0.4424!)
by substituting
for a,
3.
1952
o.
l>,
c in
we
415
V = a + bD + cD +
n-
and proceed
a
similarly,
3-1935,
we
<//><
shall find
^= +
0-4735
0.4735
c-
</
0.8563,
4-
0.0675
so that
V=
'
3.1935
O.S563/)
[H
is
0.0675
D*
given by
=0.45
z>
obtained.
to
occur
in the
measured
So
velocities.
in
questioned.
be found
in Art. 106.
where a
= 0.0036678,
and
a, b, c,
at
temperature
are constants,
and H.
We
C.
is
found
+ i in number, to
have also given that
>/
when
t
Travaujc
ft
Mtmoircs
<fu
= ioo
C.,
H = 760
Bureau international
""
lies
raids
ct
Mfsurfs,
Paris,
1881.
416
To
find the
constants
a,
/>,
c,
we proceed
H in
and
as follows:
equation
and eliminating
i,
a,
we
have
= 7OO
IO
= 760
10
\i
""
+ looo
^' + c^2 +
log //
log 760
\i+iooo
at)
i -(-
i -j-
-t)
suppose.
= bpi
fpv
(2)
t, ... by selecting
Next compute approximate values of
in Eq.
and substituting the observed values of t and
/',
tions
of the observa-
Let
2.
b'
c',
denote the values found from the solution of the resulting equations, and
//' denote the corresponding value of//, so that
If
Also, let x,
values b' c
,
then, from
y,
,
'
760
io~^>'+ c >*+-
'
let
>
(3)
to find the
(2),
log 760
log \ff'
+ (^
ff')\
log 760
log //'
(b'
+ r )/i +
remembering
= b'p +
l
c'p-i
'
that,
from
(3),
we have
The
solution
may
be finished
in
the
usual way,
forward.
417
insufficient or contradictory
is
is
method of
the
nomena
phenomenon
yl/,,
<f -j-
J/
a,
(n
41 8
&
val 6
that
is,
-.
n
360
is
vari-
able
= X+ h,
f(<p)
sin (a, -f
<p )
-f
/t t
sin (a,
2<p)
1?
the
of
values
assigned
variable
<p.
Substitute
M^,
lt
- i)0
26,
(;/
for/(<p), and <p, <p -f 6, <p
<p
in the above equation, and we shall have n equations
for
from which to determine a number of constants not exceed3
.Mu
<f>,
ing n
i.
To
lighten the numerical work let us assume the arithmetic mean of the observed values as an approximate value
of X, and
Then
if,
let
x denote
as usual,
is
placed
may be expressed by
the
general formula
-4-
/i,
sin (,-(-
//
-f-
where
<f>
assumes
Writing them
x -\-y, cos 6 -f s
in &)
sin (a,
all
+ 2<p+2mff)+
values from o to n
-/m
,,
at full length,
sin
sin
J
6 -(- j
-f-
cos 2 o -f - 2 sin 2 o
cos 2 S-\- s, sin 2 S
+
-|-
/,
=v
?;,
t ,
vy
=
z =/
h, sin (a,
j,
cos
tl
419
(a,
-f
4-
=
=
j,
<p),
s,
<p),
//,
+
+
sin (a,
//,
cos
(a,
2<p),
2<p),
Art. 9)
= [/] = o
nx
I^^co
-j,
ft
= 2*4
-^
-2,= 2/m +
where
has
values from o to n
all
sin
sin
2m 6
i.
if
spond to
o,
simpler to
6,
26,
M Mv M
lt
...(
1)6,
of f(<f) corre-
where
116
= 360,
of
first
all
it
in
is
the
form
X-\-y, cos
f((f>)
-f-
5, sin
<p -\-
-|~
z,
"
cos
2<p -f-
sm 2 ^ +
where
^r,
Then with
54
=h
=
//,
sin a,
COS
a,
j,
<cr
=
=
//,
sin
a,,
//,
COS
...
...
we
420
of the
Hence
the function
Two
195.
(a)
known.
is
10,
T<y>-
36
5 *,
5
)< 2
5 s2
7)
(A.
/,)
(/.
36
72
It will
written
36
72
- /) - (A - /.)
{(/- /) + (A
/.)
(/,
(/s
{(/
cos 72
sin 72
A) \ cos 36
+ A)-(A +
/.)}sin 36
A.)
(A
6=30
n=i2,
(b)
= (A - A) +
62,= (/
6>< 2
6s 2
= (A
=
(/a
- /.) -
(/.
9)
-A
i) }
sin 30
i) (
8)
2)
(/.,
The
is
's
in
each parenthesis
The m.
196. The Precision.
found from the usual formula
where n
is
the
s. e.
is
Check
of [w].
M=
Now
421
[//]-
\hi
i~i
r/-/ 1~\*
[<:<:.2]
[V/.2],
their values
from
-jr. i. The mean monthly heights of the water in Lake Michigan at Chicago below the mean level of the lake from iS6o to 1875, for the 12 months of
the year 1868, were as in column
of the following table:
422
A=
Let
= the
<p
The correction
10 the first
sin(j
h\
reading will be
to the
hi sin((iri
tp)
//a
sin (a 2
A +
(p)
hi sin
(nr 2
+ 2A +
2/ii
Suppose now
cos(o' 1
A
--h
A
q>)
been read on by
+ &,
q>, tp
amount by which
We
A=
M+
2/ii
cos
(cti
q>
tt
-\
If
the
x +
assumes
we
all
cos
(IT,
values from o to n
assumes
Expanding
all
of the
different parts
graduated dis-
&=-
+ m&) sin
.
<p
2cp
A+
2mf-)) sin
i.
+
where
+ zcp)sinA
A
4 ---h
scale distance is
A
A
--h q> + ;)sin
as an
take
i)Q
(n
n of the
2cp)
+ A
h 2^ 2 cos (a 2
sin
the
M of the
A =M+
2<p)
second reading
2/i z
tlie
general formula
A
2
cos
(o
values from o to n
A+
-f
2<p
+ 2m@) sin A +
may be completed
as in Art. 194.
The screw of the filar micrometer of the 26-in. refractor of the U. S. Naval
Observatory was examined for periodic error by'Prof. Hall in 1880 (Washington Observations, 1877).
423
The micrometer was placed under the Harkness dividing engine, and the
of a revolution of the screw was measured
distance corresponding to each
means
for
each
'
of a revolution
Microm.
to the engine.
The following
are the
424
and
to the
reading of microscope
hi sin
The
correction to the
h?.
and
a?
sin (a<i
2<p)
<p)
2hi cos
(oti
h^ sin
+A
+-
A* sin
to the
A = MI +
the
mean reading on
(A +
+ hi sin (a? +
<p)
mean reading on
h* sin
(i +
writing 180
2,
first
(<T 4
...
2q>)
signal
is,
+ 4<p) +
therefore,
<T 4
4 (A
observed value
<p)
MI
we have
2(p) sin
+ 2/i
cos (a 4
A + 4^)
sin
zA
Suppose now, sighting at the same two signals, that readings have been made
n different parts of the circle, so that to tlie observed values of the angle
at
the
solution as in Ex.
C-).
tp
(n
i)0,
Complete the
2.
425
cos o
yi cos 60
y-t
cos 300+
The normal
zi sin
-f 2 a
sin 60
z a sin
z 4 sin
120=
j4cos6oo +
z^ sin
600=: + 4.06
300+
3Z
3^4
Hence
sin
z\
=
=
=3^4 =
4-22
11.44
1-04
0.55
1.03
2.23
equations,
3>'i
instrument
= +
+ y* cos o +
+ y^ cos 120 +
measured with
this
given by
+ A 4
2<p)sin,4
i".
54 cos (91
zA +
49)) sin
zA
venet,
APPENDIX
I.
HISTORICAL NOTE.
as
198. The first account of the method of least squares
now employed was published by Legendre in 1805, in his
was published
in
measures
To
To
correct a survey.
Gauss published
There
is
in 1809, in
a copy of the Analyst containing Adrain's proofs in the library of the American
An account of Adrain's life will be found in the Demo-
55
APPENDIX.
428
veloped the subject in its principles and practical applicathat comparatively little has been added by later
An English translation of the TJieoria motus by
writers.
Admiral Davis, U. S. N., was published in 1858, and a
French translation of Gauss' memoirs on least squares by
Bertrand in 1855.
The main contributions to the subject, aside from those
mentioned, have been made by Laplace in the fundamental
principles, and by Bessel, Hansen, and Andrse in its applications to astronomical and geodetic work.
199. Many proofs of the law of error have been given.
tions
1809 assumes that if a series of obsermade of a quantity, the arithof the observed values is the most probable
in
metic mean
that the
made
suming that positive and negative errors are equally probable and that the number of observations is infinitely great.
form of
APPENDIX.
and Tail,
429
i.
Clarke,
Geodesy.
referred
is
entitled
List of
Squares,
New
Haven,
1877.
APPENDIX
II.
It
a quantity, T,
observed value of
we should have
is
T, then,
we make
to be determined,
if
But
since,
may
if
and
J/, is
an
we
we
where
t ,
M^
J,,
a,
Now, an
Thus
in
APPENDIX.
430
infinitesimal,
and
generally
,,
results,
if
i
and
-\-
.1-
are
-f-,
can occur
this
and n
in
2n
-
are
(211
i)
12.
the error
(n
(n
-4-
111
2we
-\-
i )
m)
Hence
the numbers expressing the relative frequency of the errors (that is, the number of times they may
be expected to occur) are equal to the coefficients in the
th
development of the 2n power of any binomial.
ways.*
The element
errors, infinite in
number, being
infinitely
from o
to 2ns, the
maximum
* Sec
error.
If,
therefore,
J denotes
APPENDIX.
431
= 2ine
= (2m -f 2)f
J
dJ
and therefore
J
Calling / the relative frequenc) of the error
that of the consecutive error J -J- */J, we have
7
2n(2n
1}
2(2-
I)
ancl/-f-
df
(n -\-m-\-i)
J,
;//
1
...
(//-f
w + 2)
Hence, by division,
n -\-m-\-\
/"
or
2W-J-
f//"_
~
Now, since dA
may write
is
infinitely small in
df_
7
Also, since ^/"
is
in
comparison with
J,
J,
comparison with /, 2J is
and we may neglect J in the de-
df_
we
2J
infinitely small in
nominator
~ 2J_
We
have, therefore,
APPENDIX.
432
we have
*--2h*AdA
Integrating and denoting the value of/,
The
o,
dA,
when J
o,
by/
between A and A
-j-
dA, and
If
denominator the
we denote
total
The
Suppose
function
num-
this probability
by
for the
T of a
certain
known
T = f(X,
form,
F,
.)
APPENDIX.
in
433
MM
Y,
unknowns
are the
for
to be
If
7\
second,
values of X, F,
therefore of
., and
for the probabilities, before the first,
.
between
lie
expected
respectively, are
where
c lt cv
//,,
//
The
are constants.
probability </>
of these errors, which are independent of each other,
given by (Art.
But
If
4>
now
J,,
M^MV
J,, 4,,
.,
</>,
are assumed,
arbitrary values of X, Y.
will receive values, and therefore a value of
.
is
5)
true values of T, X, Y,
are unknown.
T,
all
will
will
t ,
F,,
Substitute
APPENDIX.
434
V F
s,
we have no longer
T T
t
iy
M^
T^
Then
.
.,
F,
} ,
.,
comes
would be
and the most probable set of values of X, Y,
that which corresponds to the maximum value of this expression, which can only happen when
.
[/^V]
same
for the
is
set of values of
minimum
X,
F,
This
is
202.
Young,
for other
Philosophical Transac-
WahrscJieinlicJi-
mal
Calculus, vol.
actions, 1865
ii.,
Oxford, 1865
Kummell, Analyst,
Tait,
vol.
iii.,
Price, Infinitesi-
APPENDIX.
TABLE
Values
a
r
of
0/
I.
435
APPENDIX.
TABLE
II.
APPENDIX.
TABLE
Factors for Peters
III.
Probable-Error Formulas.
437
DC
Los Angeles
9 ^J.6C
*
i DUE on the last date stamped below.
TMPSookk is
IBJECT TO FINE
IF
EDUCATION
RECEIVED
SEP
SP
2 2 1970
291972
Form L9-116i-8,'62(D1237s8)444
UCU-ED/PSYCH
QA
Library
275 W93t
L 005 647
72 5
EDUCATION
LIBRARY
QA
275
W93t