01.03.2013 Views

Applied Statistics Using SPSS, STATISTICA, MATLAB and R

Applied Statistics Using SPSS, STATISTICA, MATLAB and R

Applied Statistics Using SPSS, STATISTICA, MATLAB and R

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

368 9 Survival Analysis<br />

The exponential model can be fitted to the data using a maximum likelihood<br />

procedure (see Appendix C). Concretely, let the data consist of n survival times, t1,<br />

t2, …, tn, of which r are death times <strong>and</strong> n – r are censored times. Then, the<br />

likelihood function is:<br />

n<br />

−λti<br />

δi<br />

−λt<br />

L e e i 1−δ<br />

⎧0<br />

ith<br />

individual is censored<br />

( λ)<br />

= λ<br />

i<br />

∏ ( ) ( ) with δ i = ⎨<br />

. 9.25<br />

i=<br />

1<br />

⎩1<br />

otherwise<br />

Equivalently:<br />

n<br />

∏<br />

i=<br />

1<br />

δi<br />

−λti<br />

L(<br />

λ)<br />

= λ e , 9.26<br />

from where the following log-likelihood formula is derived:<br />

n<br />

∑δilog λ − λ∑<br />

n<br />

log L(<br />

λ ) =<br />

t = r log λ − λ t . 9.27<br />

i<br />

i=<br />

1 i=<br />

1<br />

n<br />

∑<br />

i<br />

i=<br />

1<br />

The maximum log-likelihood is obtained by setting to zero the derivative of<br />

9.27, yielding the following estimate of the parameter λ:<br />

n<br />

= ∑ i=<br />

i t ˆλ<br />

1<br />

. 9.28<br />

1 r<br />

The st<strong>and</strong>ard error of this estimate is ˆ λ / r.<br />

The following statistics are easily<br />

derived from 9.24:<br />

ˆ<br />

0 . 5 =<br />

λˆ t ln 2 / . 9.29a<br />

λˆ ˆ = ln( 1/(<br />

1−<br />

p))<br />

/ . 9.29b<br />

t p<br />

The st<strong>and</strong>ard error of these estimates is tˆ p / r .<br />

Example 9.8<br />

Q: Consider the survival data of Example 9.5 (Heart Valve dataset). Determine<br />

the exponential estimate of the survivor function <strong>and</strong> assess the validity of the<br />

model. What are the 95% confidence intervals of the parameter λ <strong>and</strong> of the<br />

median time until an event occurs?<br />

A: <strong>Using</strong> <strong>STATISTICA</strong>, we obtain the survival <strong>and</strong> hazard functions estimates<br />

shown in Figure 9.7. <strong>STATISTICA</strong> uses a weighted least square estimate of the<br />

model function instead of the log-likelihood procedure. The exponential model fit<br />

shown in Figure 9.7 is obtained using weights nihi, where ni is the number of<br />

observations at risk in interval i of width hi. Note that the life-table estimate of the<br />

hazard function is suggestive of a constant behaviour. The chi-square goodness of<br />

fit test yields an observed significance of 0.59; thus, there is no evidence leading to<br />

the rejection of the null, goodness of fit, hypothesis.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!