Instrumental variable and Cox proportional hazards model

[Descartes]

Dear all,

I want to use an instrumental variable in survival analysis, and I think the PROC SYSLIN is only for linear or logistic regression? Is there command specifically for Cox proportional hazards model?

I am thinking to use below method, is it correct? Thanks for your help!

PROC LOGISTIC DATA= data1;

  MODEL GRP = Distance;

  OUTPUT OUT = data2

  PREDICTED=IV_grp;

RUN;

PROC PHREG DATA =data2;

   CLASS  AGE RACE INCOME EDUCATION / PARAM=REF REF=First ;

   MODEL TIME*Status(0) = IV_grp AGE RACE INCOME EDUCATION ;

RUN;

[SteveDenham · Accepted solution]

I think you may want to include the additional variables (AGE RACE INCOME EDUCATION) in the PROC LOGISTIC fit, as distances may cluster somewhat for these variables.

Steve Denham

[Descartes]

Thanks Steve! Is there any test I can do on this instrumental variable?

[SteveDenham]

Well, there should be a Wald chi-square test in the output for PROC PHREG for this variable.

Steve Denham