Dear Statalist,
I am attempting to estimate 5 year survival after Cox proportional hazards regression with a multiply imputed dataset.
My issue is with prediction of baseline survival;
I am trying to use mi predictnl as basesurv is not a linear prediction (having tried & failed using mi predict)
[CODE]
mi stset death_censor_date, failure(death==1) origin(diagnosis_date) scale(30.4) id(ID)
mi estimate, saving (myestfile, replace): stcox var1 var2 var3
predict xb using myestfile, xb
mi predictnl basesurv = predict(basesurv) using myestfile
Error message:
e(sample) not found
predict(basesurv) invalid
I would be so grateful if anybody had any suggestions.
Thank you
I am attempting to estimate 5 year survival after Cox proportional hazards regression with a multiply imputed dataset.
My issue is with prediction of baseline survival;
I am trying to use mi predictnl as basesurv is not a linear prediction (having tried & failed using mi predict)
[CODE]
mi stset death_censor_date, failure(death==1) origin(diagnosis_date) scale(30.4) id(ID)
mi estimate, saving (myestfile, replace): stcox var1 var2 var3
predict xb using myestfile, xb
mi predictnl basesurv = predict(basesurv) using myestfile
Error message:
e(sample) not found
predict(basesurv) invalid
I would be so grateful if anybody had any suggestions.
Thank you
