12-22-2015 11:09 AM
With the hazardratio statement you can get phreg to compute hazardratios for the parameters of interest. For example, if your parameter of intersest is "A" and you have an interaction term with "A*B", then you can with the hazardratio statement you can get the effect of A for each value of B. Like this example:
proc phreg data=mydata ; class A B/param=ref; model (entry exit)*event_(0)=A*B; hazardratio A/at(B=all); run;
The hazardratio statement does not decide what model you use. It is only used to specify what output you will see from the fitted model.
The strata statement is used to specify if the background hazard rate should be dependent on some class variable. If you have a strata variable you in principle adjust for an interaction term between the underlying time scale and the strata variable. You can not have you covariate of interest in strata, because it will not give you estimate for strata effects - the background hazard rate is non-parametric. A very common used model in epidemiology is to have birthyear as strata and age as underlying time, since you then effectively also have a non-parametric adjustement for calendar time.
12-23-2015 06:42 AM
My apologies if I was not clear in my intial post. I meant to ask about 'Hazard Ratio' in output window which we used to get after the successful completion proc phreg.