topic survival analysis with multiple imputation in Statistical Procedures

survival analysis with multiple imputation

smoore3790 — Tue, 18 Apr 2023 20:30:12 GMT

I am using SAS to do a multiple imputation, and I am having trouble figuring out proc MI. I am looking at a comparison of a control with 126 observations to the cases with 63 observations. The missing data is specifically only in the PercentUsed variable. This is what I have so far. Really appreciate any help.

proc mi data=FinalMod nimpute=126 seed=12345 out=imputed;
var days case sleep exercise age PercentUsed  ;  
run;

PROC PHREG data= FinalMod;
MODEL days*case(0) =  sleep exercise age PercentUsed/ RISKLIMITS=BOTH
TYPE3(ALL);
RUN;

proc mianalyze data=imputed;
  modeleffects est=phreg;
run;

I am expecting results that look similar to the original proc phreg, but slightly different for the PercentUsed variable

Re: survival analysis with multiple imputation

SAS_Rob — Thu, 20 Apr 2023 14:15:19 GMT

It isn't clear exactly what your question is. You will not get any output from the code you sent, at least with respect to the multiple imputation modeling and combining of estimates.

The PHREG step is on the wrong data set and needs to be done BY _IMPUTATION_ and the syntax in the MIANALYZE step is not valid. You should run this instead:

PROC PHREG data= imputed;
by _imputation_;
MODEL days*case(0) =  sleep exercise age PercentUsed/ RISKLIMITS=BOTH
TYPE3(ALL);
ods output ParameterEstimates=phreg_parms:
RUN;

proc mianalyze parms=phreg_parms:
  modeleffects sleep exercise age PercentUsed;
run;

Re: survival analysis with multiple imputation

smoore3790 — Thu, 20 Apr 2023 17:08:31 GMT

Thanks so much for the help, I really appreciate it. This worked perfectly and completely fixed my issue!!!