Hi, I would like to apply Phreg procedure to test whether two groups of crayfish differ in their eagerness to leave shelters. I measured this over 30 minutes, but I my additional hypothesis is that this response will be seen over the first 15 minutes, and it will disappear afterwards. So, I included in my model the effects of "group" variable, but also an interaction between "group" and a time-dependent covariate that takes a value 0 during the first 15 minutes and a value 1 during the last 15 minutes:
model time*censored(0)=gr gr*status;
if t<15 then status=0;
My question is whether I can use Cox model in this situation. I want to test for the significance of this interaction and if it is significant, my scientific hypothesis will be accepted then, but at the same time in a classical approach the Cox model requires assumption about nonsignificance of such interaction (as far as I understand). I would be grateful for any opinion on my problem.