Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • mi estimate using - can one access returned results from individual imputation sets?

    Hello

    I was wondering whether it is was possible to access the returned results (specifically the matrices) for each individual imputation after having run -mi estimate- with save(miest) option? I've scoured the reference manual and I couldn't see how it could be done even though the individual results must be in 'miest.ster' somehow.

    With -mi estimate using- one can only get the pooled mi estimate results returned, using whichever imputations are chosen, but you have to select more than one imputation set. One can type -est use miest.ster- in the standard manner but typing -eret list- suggests this brings up the stored results from the 1st imputation only.

    If not, is it to prevent the implementation of dubious post estimation procedures in an MI setting?

    thank you
    Matthew

  • #2
    Hi, Matthew.

    Why don't you run the usual command in the ith imputed dataset?

    Like:

    regress y x1 x2 if _mi_m==2

    All the best,

    Tiago


    Comment


    • #3
      see "help mi_xeq"

      Comment


      • #4
        See

        Code:
        help estimates
        You will find that

        Code:
        estimates describe using miest.ster
        will return in r(nestresults) the number of saved results which will be M+1, where the first M datasets correspond to the imputed datasets and the last set of results holds the combined MI estimates.

        You will further see that you can access other results than the first set typing

        Code:
        estimates use miest.ster , number(#)
        Best
        Daniel

        Comment


        • #5
          Thank you Daniel - that is exactly what I was looking for. When I say I scoured the manual, I admit I was only looking in the MI section.

          @Tiago, the reason I didn't want to have to run individual estimations on the imputation sets again was because each one is taking 3 hours or so, and I have 20 imputation sets!

          Comment

          Working...
          X