topic Cox Regression: Can you get hazard ratios for an interaction term? in Statistical Procedures

Cox Regression: Can you get hazard ratios for an interaction term?

clcominsky — Tue, 30 Aug 2016 16:34:06 GMT

When reporting hazard ratios for Cox regression analysis, is it common to report the hazard ratio for the interaction term itself?

For example, I have a model with 3 terms:

a*b

Using hazard ratio statements in SAS 9.4, I get a hazard ratio for 1) a at the mean of b, and 2) b at the mean of a. My understanding is that these hazard ratios are hazard ratios for the main effect variable (variable a or b) while holding the interacting variable constant. Is my understanding correct?

Is it possible to get a hazard ratio for the interaction term?

Thank you

Re: Cox Regression: Can you get hazard ratios for an interaction term?

JacobSimonsen — Tue, 30 Aug 2016 21:07:55 GMT

Yes, you can include interaction-terms. You should use the option "/param=glm" in the class statement.

With the hazard-ratio statement you can easily get the effect of one of the terms on each level of the other term. Like this:

data simulation;
*the the hazard ratio for A=1 vs A=2 is 1, 1.5 and 2 for respectively B=1, 2, 3;
  array a_effect{3} _temporary_ (1,1.5,2);
     do B=1 to 3;
  do A=1 to 2;
	 do i=1 to 100;
	   t=rand('exponential',1/exp((a=1)*log(a_effect[B])));
	   output;
	 end;
	 end;
  end;
run;
proc phreg data=simulation;
  class a b/param=glm;
  model t=a*b;
  hazardratio a/at(B=all);
run;

Good luck.

Re: Cox Regression: Can you get hazard ratios for an interaction term?

clcominsky — Tue, 06 Sep 2016 18:34:20 GMT

Thank you, Jacob. Your response was very helpful. Thank you for taking the time to provide it.

--Cynthia