What does type 3 in Cox model (proc phreg) with random effect means

Matyas — Tue, 12 Jul 2022 14:44:01 GMT

Hi,

I am running the Cox model with random effect (simple random effect of study site) as follows:

proc phreg data=mydata;
  class study fixed_covariate; 
  model time*cnsr(1) = fixed_covariate;
  random study / solution;
  hazardratio fixed_covariate/diff=ref;
run;

As part of results I got Type 3 Anova model table:

I understand that fixed_covariate has no effect on survival, but may I say that there is a significant difference between study sites on survival? What does this adjusted p-value for random effect stand for?

Thanks for any advice.

Re: What does type 3 in Cox model (proc phreg) with random effect means

svh — Wed, 13 Jul 2022 14:47:53 GMT

I think you can interpret that p value to mean that there is a statistically significant effect of the clustering variable in this shared frailty model. SAS does have a worked example of a similar type of analysis:

https://support.sas.com/documentation/cdl/en/statug/63962/HTML/default/viewer.htm#statug_phreg_sect056.htm

topic What does type 3 in Cox model (proc phreg) with random effect means in Statistical Procedures

What does type 3 in Cox model (proc phreg) with random effect means

Re: What does type 3 in Cox model (proc phreg) with random effect means