Nonparametric Methods in Statistics with SAS Applications

134 TIME-TO-EVENT ANALYSIS

• thickness: For a one millimeter increase in melanoma thickness, the hazard in-

creases by 100(exp{0.61677}−1)% = 85.29%.

This model may be used for prediction. For instance, the estimated relative risk for

a 50-year-old female with a melanoma on the upper arm that is 2 mm thick is com-

puted as



r = exp



0.04712(50−58.9) −1.20811(0 −0.4)+0.65729(0 −0.53333)+

0.61677(2 −3.03667)



= 0.39611. For this patient, the probability to survive for 5

years is estimated as



S(5) = 0.50185

0.39611

= 0.7610.

The same model may be fit by using the CLASS statement. The code is

proc phreg data=melanoma outest=betas;

class gender (ref=’F’) site/param=ref;

/* to make ’F’ a reference as in the above model*/

model length*censored(1)=age gender site