08-11-2016 04:03 PM
I have a question regarding how to determine model adequacy for cox-proportional hazard model using counting process style of input. In the regular style of input, I know I can use assess ph option in proc phreg to determine whether the model fails the non-proproportionality assumption. I did this prior and determined some covariates were time dependent. Therefore we have changed the covariates to time dependent which lead to the creation of the dataset with the counting process style. However, for the counting process style of input, the assess ph option does not work. I am wondering
1) Do I need to test non-proportionality in this case
2) If yes to 1) then what options can I use.