BSTA 6652 HW#1 Answer Key– Winter, 2012

Ex. 2.2, 2.4, 2.6, 2.10(a)(b)

2.2 (a) Let X denote the time (days) to tumor development.

The survival function of the Weibull distribution with α = 2 and λ = 0.001 (using Table 2.2):

The probability that a rat will be tumor free at 30 days (in other words, the time the rat develops
a tumor is greater than 30 days) is

For x = 45 days,

For x = 60 days,

2.2 (b) From Table 2.2, the mean time to tumor with α = 2 and λ = 0.001 is

Note that . Then,

2.2 (c) Hazard rate function:

For x = 30 days,

For x = 45 days,

1
 BSTA 6652 HW#1 Answer Key– Winter, 2012

For x = 60 days,

2.2 (d) The probability of survival, 1 – p, at time xp is

Solving for xp,

Therefore, the median time to tumor is

2.4 (a) Show that the hazard rate has a bathtub shape and find the time at which the hazard rate
changes from decreasing to increasing.

The hazard rate for the exponential power distribution is

For α = 0.5,

The hazard rate for various values of λ is plotted as follows. The hazard rate has a bathtub shape
for positive λ. Here are several examples:

2
 BSTA 6652 HW#1 Answer Key– Winter, 2012

Hazard rate function for

exponential power distribution

5
λ=2
λ=1

4
λ = 0.5

3
h(x)

2
1
0
0 2 4 6 8

For x < 1 / λ , λ x − 1 < 0, so , i.e. is decreasing here. For x > 1 / λ , λx − 1 > 0, so

, i.e. is increasing here.

To find the time at which the hazard rate changes from decreasing to increasing for α<1, find the
local minimum:

For α = 0.5,

2.4 (b) If α = 2, show that the derivative of the hazard rate is positive for all x > 0.

Since ≥ 0 always, the other quantity in square brackets must be shown to be positive
for all x ≥ 0. For α = 2,

3
 BSTA 6652 HW#1 Answer Key– Winter, 2012

which is positive for all x ≥ 0. Therefore, the hazard rate of x is monotone increasing.

2.6 (a) Let X denote the time to death (in months) of a randomly chosen mouse. The survival
function of the Gompertz distribution:

For θ = 0.01 and α = 0.25,

The probability that a randomly chosen mouse will leave at least one year is the probability it is
still alive at one year (12 months):

2.6 (b) The probability that a randomly chosen mouse will die within the first six months is

2.6 (c) The probability of survival, 1 – p, at time xp is

Solving for xp,

Therefore, the median time to tumor is

2.10 (a) The piecewise, constant hazard rate model.

4
 BSTA 6652 HW#1 Answer Key– Winter, 2012

On the interval [0, τ1), the survival function is defined as

On the interval [τ1, τ2), the survival function is defined as

On the interval [τk–2, τk–1), the survival function is defined as

On the interval [τk–1, ∞), the survival function is defined as

All together, the survival function for this model is

In short, for (

For x = 0, the probability of survival is 1, as expected. As x tends to infinity, the probability of

survival tends to zero. This function is also piecewise continuous.

2.10 (b) The mean residual-life function.

5
 BSTA 6652 HW#1 Answer Key– Winter, 2012

On the interval [0, τ1), the mean residual-life function is defined as

On the interval [τ1, τ2), the mean residual-life function is defined as

On the interval [τk-1, ∞), the mean residual-life function is defined as

All together, the mean residual-life function is

2.10 (c) The median residual-life function.

6
 BSTA 6652 HW#1 Answer Key– Winter, 2012

The median residual lifetime is the length of the interval from time x to the time where one-half
of the individuals alive at time x will still be alive. The median residual lifetime is mrl such that

The value of mrl depends on x. Here is an example of how to find mrl. Let x=0, then ,
the median life time. First of all, find the value of

Second, find the interval [τj-1, τj) such that . Then falls into the interval [τj-1, τj)
and Solving for gives
.

Problem 1. Given the lifetime variable Y is exponentially distributed with a positive constant β ,
-y/β
we know that S(y) = e for y ≥ 0. Therefore for t ≥ 0 we can write:

-(y+t)/β -y/β -t/β

S(y+t) = e =e *e = S(y) * S(t). Thus a lifetime variable that is

exponentially distributed exhibits the memoryless property: S(y+t) = S(y)S(t), y≥0, t≥0