01-05-2017 01:18 PM
I am performing survival analysis for a study and want to stratify a couple of variables that did not meet the Proportional Hazards Assumption so that I could fit a Stratified Cox model. Normally, you would use the STRATA statement in PROC PHREG to fit different baseline hazard functions in different strata of each covariate that did not satisfy this assumption. Since the STRATA statement in SURVEYPHREG is completely different from that in PHREG, I'm not sure which statement I should use to fit a Stratified Cox model that simultaneously accounts for the study's stratified sample design and covariates that do not meet the PH assumption.
proc phreg data=x;
model survtime*conv(0)= exposure cov3 cov4 exposure*cov1 exposure*cov2 exposure*cov3 exposure*cov4; <-including all exposure*covariate interaction terms
strata cov1 cov2; <-those that do not satisfy the PH assumption
proc surveyphreg data=x;
strata sampling_stratification_variable; <-this STRATA statement is NOT the same as the STRATA in PROC PHREG!
model survtime*conv(0)= exposure cov3 cov4 exposure*cov1 exposure*cov2 exposure*cov3 exposure*cov4;
[insert a statement that stratifies on the covariates not meeting the PH assumption];