Help with mestreg after multiple imputation

[Gius DAndrea]

Hi

I'm running survival analyses (mestreg) on Stata using a dataset with 60 imputations to account for missing data.
When I run the command "mi stset time" the software will still take into account the observations with missing data on time from the unimputed dataset (m=0).

In the regression output the number of observations will drop from 4,356 to 3,960 even if I specify that I only want the model to be fitted within the datasets m=1 to m=60.

The code I use is the following:

**** begin code
capture program drop mymestreg
program mymestreg
mestreg c.month i.age i.gender i.ethnicity || site:, distribution(lognormal) time covariance(unstructured)
end

mi stset time

mi estimate, imputations(1/60) saving(miest,replace) eform cmdok: mymestreg
**** end code

Thanks for your help!

[Gius DAndrea]

A little update, hoping that someone can help. The issue seems to be that Stata will not allow imputation of the survival time. I had 2 or 3 hundreds of observations where survival times was indeed missing and thus had to impute those values. Any hint at how to solve this? I've though to run the analyses separately on each dataset but is there an easy way to pool the estimates then?

[Gius DAndrea]

Nobody?