SPSS Statistics

SPSS Statistics

Your hub for statistical analysis, data management, and data documentation. Connect, learn, and share with your peers! 

 View Only

  • 1.  How to pool multiple imputation results in version 26?

    Posted Mon June 21, 2021 02:06 PM

    Hi, our university just transitioned from version 25 to 26.  In version 25 (and earlier), multiple imputation data sets had the symbol/square in the upper right hand corner that one could click on to identify the data set as an MI data set; this appears to be gone in version 26.  I have discovered how to run regressions for each of my 100 imputed data sets using the "split file" command and "compare groups" option, but this does not produce pooled data for the 100 regressions.

    Any suggestions as to how I might produce pooled data for MI results would be greatly appreciated!

    Thanks,

    Dave



    ------------------------------
    David Bennett
    ------------------------------

    #SPSSStatistics


  • 2.  RE: How to pool multiple imputation results in version 26?

    Posted Mon June 21, 2021 03:19 PM
    I see that symbol at the right end of the edit bar area (looking at V27 and 28).  But AFAIK, this is only a convenience for looking at a particular replication dataset.  It does not affect what happens when you run a procedure.  You should get automatic MI results once the split is set.

    The split should happen automatically, but if it doesn't follow the advice offered when you open the file...

    >Note # 5268.  Command name: GET FILE 
    >This appears to be a Multiple Imputation dataset.  In addition to the original 
    >data, a multiple imputation dataset contains multiple copies of the data in 
    >which missing values have been replaced with imputed values.  If you want this 
    >dataset to be treated as a multiple imputation dataset, use the SPLIT FILE 
    >command to define the Imputation_ variable as a SPLIT FILE variable before 
    >performing analyses. 
    DATASET NAME DataSet1 WINDOW=FRONT. 
    SPLIT FILE 
     LAYERED BY Imputation_.

    You would then see "Split by Imputation_" at the end of the status bar, and the imputed values should have a colored background in the imputation splits looking in the DE.

    If this isn't happening, I  can only offer two guesses.
    1.  It isn't actually a MI dataset, or
    2.  Your copy of Statistics is not licensed for the MI option.

    --



    Original Message