Standard deviation and median of multiply-imputed survey data

[Ayalew Astatkie]

I have a multiply-imputed (M=20) dataset. I want to describe normally distributed numerical variables using mean and standard deviation. For non-normal numerical variables, I want to use the median as a descriptive summary. However Stata (version 12) gives standard error (not standard deviation) after mi estimate: svy: mean varlist. The post-estimation command estat sd doesn’t work after mi estimate. Nor could I also get a command for estimating median after the prefix commands mi estimate: svy. Would anyone suggest me commands (if any) for estimating standard deviation and median for multiply-imputed data?

[Stephen Jenkins]

How about using qreg? This can be used after mi estimate. See help mi_estimation##estimation_command

[Ayalew Astatkie]

qreg also doesn't work after mi estimate: svy.

[mcaplan]

Ayalew Astatkie -- did you ever find out the way to do this? I'm having the same issue: using multiply imputed survey data and wanting to know standard deviations of the means, or at least if I can legitimately compute the SD from the standard errors that mi estimate: svy: mean xxx can give me.

[Ayalew Astatkie]

Hi mcaplan. Sorry for being late to respond to your query. Yeah, I have got solution to my problem. Refer to it here on CV.