Professional Documents
Culture Documents
GLM
Design matrix
Effect size subject-series
Voxel-wise group analysis
Group
effect size
statistics
Subject
groupings
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
Standard-space
brain atlas
subjects
Single-subject
effect size
statistics
Single-subject
effect size
statistics
Single-subject
effect size
statistics
Single-subject
effect size
statistics
subjects
Register
subjects into
a standard
space
Effect size
statistics
Statistic Image
Significant
voxels/clusters
Contrast
Thresholding
k
+
k
First-level GLM
on Marks 4D FMRI
data set
Does the group activate on average?
0 effect size
A simple example
Group
Mark Steve Karl Will Tom Andrew
Y
k
= X
k
k
+
k
Marks
effect size
Does the group activate on average?
0 effect size
A simple example
Group
Mark Steve Karl Will Tom Andrew
Y
k
= X
k
k
+
k
Marks
within-subject
variance
Does the group activate on average?
0 effect size
A simple example
Group
Mark Steve Karl Will Tom Andrew
All rst-level GLMs
on 6 FMRI data set
Y
K
= X
K
K
+
K
Does the group activate on average?
What group mean are we after? Is it:
1. The group mean for those exact 6 subjects?
Fixed-Effects (FE) Analysis
2. The group mean for the population from which
these 6 subjects were drawn?
Mixed-Effects (ME) analysis
A simple example
Group
Mark Steve Karl Will Tom Andrew
Do these exact 6 subjects activate on average?
Fixed-Effects Analysis
Group
Mark Steve Karl Will Tom Andrew
0 effect size
g
=
1
6
6
k=1
k
estimate group effect size as
straight-forward mean
across lower-level estimates
Do these exact 6 subjects activate on average?
Fixed-Effects Analysis
Group
Mark Steve Karl Will Tom Andrew
0 effect size
Y
K
= X
K
K
+
K
K
= X
g
g
=
1
6
6
k=1
k
X
g
=
1
1
1
1
1
1
Group
mean
Do these exact 6 subjects activate on average?
Fixed-Effects Analysis
Group
Mark Steve Karl Will Tom Andrew
0 effect size
Y
K
= X
K
K
+
K
K
= X
g
g
=
1
6
6
k=1
k
X
g
=
1
1
1
1
1
1
Group
mean
Do these exact 6 subjects activate on average?
K
+
K
K
= X
g
g
Fixed Effects Analysis:
Does the group activate on average?
What group mean are we after? Is it:
1. The group mean for those exact 6 subjects?
Fixed-Effects (FE) Analysis
2. The group mean for the population from which
these 6 subjects were drawn?
Mixed-Effects (ME) analysis
A simple example
Group
Mark Steve Karl Keith Tom Andrew
0 effect size
Does the population activate on average?
Mixed-Effects Analysis
Group
Mark Steve Karl Keith Tom Andrew
0 effect size
Y
K
= X
K
K
+
K
k
Consider the distribution over the
population from which our 6
subjects were sampled:
2
g
is the between-subject variance
g
0 effect size
Does the population activate on average?
Mixed-Effects Analysis
Group
Mark Steve Karl Keith Tom Andrew
Y
K
= X
K
K
+
K
k
Consider the distribution over the
population from which our 6
subjects were sampled:
2
g
is the between-subject variance
g
Does the population activate on average?
Mixed-Effects Analysis
Group
Mark Steve Karl Keith Tom Andrew
Y
K
= X
K
K
+
K
0 effect size
K
= X
g
g
+
g
X
g
=
1
1
1
1
1
1
Population
mean
between-
subject
variation
Does the population activate on average?
Mixed-Effects Analysis
Group
Mark Steve Karl Keith Tom Andrew
Y
K
= X
K
K
+
K
0 effect size
K
= X
g
g
+
g
X
g
=
1
1
1
1
1
1
Population
mean
between-
subject
variation
Does the population activate on average?
K
+
K
Mixed-Effects Analysis:
K
= X
g
g
+
g
All-in-One Approach
computationally demanding
At each level:
pass these up
Default is:
FLAME1+2:
estimate mean
estimate std-error
(FE or ME)
test signicance of
mean > 0
>0?
0
s
u
b
j
e
c
t
effect size
Single Group Average
estimate mean
estimate std-error
(FE or ME)
test signicance of
mean > 0
>0?
0
s
u
b
j
e
c
t
effect size
Single Group Average
Does the group activate on average?
0
s
u
b
j
e
c
t
effect size
Single Group Average
Does the group activate on average?
0
s
u
b
j
e
c
t
effect size
Single Group Average
Does the group activate on average?
0
s
u
b
j
e
c
t
effect size
Single Group Average
Does the group activate on average?
Unpaired Two-Group Difference
estimate means
estimate std-errors
(FE or ME)
test signicance of
difference in means
>0?
0
s
u
b
j
e
c
t
effect size
Unpaired Two-Group Difference
estimate means
estimate std-errors
(FE or ME)
test signicance of
difference in means
>0?
0
s
u
b
j
e
c
t
effect size
Unpaired Two-Group Difference
Is there a signicant group difference?
0
s
u
b
j
e
c
t
effect size
Unpaired Two-Group Difference
Is there a signicant group difference?
0
s
u
b
j
e
c
t
effect size
Unpaired Two-Group Difference
Is there a signicant group difference?
0
s
u
b
j
e
c
t
effect size
Unpaired Two-Group Difference
Is there a signicant group difference?
0
s
u
b
j
e
c
t
effect size
Unpaired Two-Group Difference
Is there a signicant group difference?
Unpaired Two-Group Difference
Is there a signicant group difference?
Paired T-Test
use covariates to
explain variation
estimate mean
estimate std-error
(FE or ME)
0
s
u
b
j
e
c
t
effect size
Single Group Average & Covariates
use covariates to
explain variation
estimate mean
estimate std-error
(FE or ME)
0
s
u
b
j
e
c
t
effect size
Single Group Average & Covariates
fast RT slow
use covariates to
explain variation
estimate mean
estimate std-error
(FE or ME)
0
s
u
b
j
e
c
t
effect size
Single Group Average & Covariates
fast RT slow
use covariates to
explain variation
estimate mean
estimate std-error
(FE or ME)
0
s
u
b
j
e
c
t
effect size
Single Group Average & Covariates
fast RT slow
use covariates to
explain variation
estimate mean
estimate std-error
(FE or ME)
0
s
u
b
j
e
c
t
effect size
Single Group Average & Covariates
fast RT slow
use covariates to
explain variation
estimate mean
estimate std-error
(FE or ME)
0
s
u
b
j
e
c
t
effect size
Single Group Average & Covariates
fast RT slow
Does the group activate on average?
3 groups of subjects
Group F-tests
3 groups of subjects
Is any of the groups activating on average?
ANOVA: 1-factor 4-levels
ANOVA: 1-factor 4-levels
Fixed-Effects ANOVA:
ANOVA: 3-factor 2-levels
Fixed-Effects ANOVA:
First-level analysis:
Understanding FEAT dirs
Second-level analysis:
Thats all folks