how to use stcurve for Cox model after multiple imputation?

[Weidi Qin]

Hello, Statalist!

I have a question about the stcurve after a multiple imputation.

So, after the following command:
mi estimate:stcox var1 var2 var3

I wanted to plot the survival probability, and used stcurve command:

stcurve, survival at1(var1=1) at2(var1=2) yla(,nogrid)

but I got an error code:
last estimates not found
r(301);

My question is: how to run stcurve after multiple imputations?

[Carlo Lazzaro]

Weidi:
googling with -stcox- amd -multiple imputation- did not give back any useful entry for your problem.

[Rich Goldstein]

because MI makes several estimates (one for each imputed data set), and then combines them using Rubin's rules, any post-estimation command that requires information from the 'last estimate" is bound to fail; there are ways around it and I show one in #6 in https://www.statalist.org/forums/for...ng-mi-estimate - but you will have to come up with your own estimate of what you want to graph and then, emulating Rubin's rules as in the cited post, get what you want to graph

[Weidi Qin]

Weidi:
googling with -stcox- amd -multiple imputation- did not give back any useful entry for your problem.

Thanks, Carlo. I've googled quite extensively before posting my question here. It seems that -stcurve- cannot be performed after -mi estimate-.

[Weidi Qin]

because MI makes several estimates (one for each imputed data set), and then combines them using Rubin's rules, any post-estimation command that requires information from the 'last estimate" is bound to fail; there are ways around it and I show one in #6 in https://www.statalist.org/forums/for...ng-mi-estimate - but you will have to come up with your own estimate of what you want to graph and then, emulating Rubin's rules as in the cited post, get what you want to graph

Thank you for sharing!