Proportional hazards check for Cox regression with categorical variables?

TiaT — Sun, 04 Jul 2021 17:46:03 GMT

Hi all,

I usually check the proportional hazards assumption for Cox regression by examining schoenfeld residuals and with the ASSESS statement. My covariates are all categorical (population and temperature treatments).

Is it correct to NOT include a class statement during assumption checking? If I include a CLASS statement, the output is broken down by each level of categorical variable. If I don't, I get an output with the 2 main covariates.

Code with no CLASS:

PROC PHREG DATA=Tia.eco18;

model Day*Status(2)=Pop_Code Temp_Code / ties=exact;

OUTPUT OUT=b RESSCH=SCHPop_Code SCHTemp_Code;

RUN;

DATA c;

SET b;

lDAYS=LOG(Day);

DAYS2=Day**2;

PROC CORR;

VAR Day LDAYS DAYS2 SCHPop_Code SCHTemp_Code;

RUN;

topic Proportional hazards check for Cox regression with categorical variables? in Statistical Procedures

Proportional hazards check for Cox regression with categorical variables?