[daniel klein]

You should probably start a new thread/topic because this question has very little to do with what we have discussed so far. Anyway, you typically do not want to collapse (or otherwise summarize) multiply imputed values because doing so is removing the between-imputation variance which is the heart of MI. If you remove the between-imputation variance before doing any analyses, then there is really no point in doing MI at all.

[Mohammad Mansour]

Thank you and sorry for posting this here. Regards