09-01-2016 12:16 PM
Is it possible to use a continuous variable (x) instead of time in the proc lifetest procedure to model the difference of event rates between treatment arms with increasing variable x...and then use the log-rank p-value?
09-01-2016 01:07 PM
It is not possible with proc lifetest.
Instead you can use proc phreg, which do a Cox regression. Here you can include a contionous variable as a predictor.
The scoretest in a Cox-regression is equivalent to a logrank test when you have a class variable, so in some sense you can say the score test generalize the logrank test since it can be used also with continous variables. It is very easy to get out of phreg. Here a simple example:
data simulation; do i=1 to 1000; y=rand('normal',0,1); t=rand('exponential',10/exp(y)); output; end; run; proc phreg data=simulation; model t=y/type3(all); run;