SPSS Statistics

Expand all | Collapse all

Trouble getting pooled values

  • 1.  Trouble getting pooled values

    Posted Fri August 20, 2021 10:14 AM
    Hello!

    I have an imputed data set that I have imported into SPSS.  The imputation variable has been properly named, and SPSS recognizes it as an imputed variable.  However, I am not able to get a pooled value for the imputed variables - I only get 20 different estimates.  It's a huge data set, so I am not able to manually mark the imputed values as such.  I've tried to use pivot tables from descriptives, frequencies and regressions, but do not have the option of selecting a pooled estimate.  I also don't see an option for designating the variables as imputed using the "view" function, which was another recommendation from IBM. 

    Can anyone help?  I've included an example below - the dropdown does not give me the option of a pooled estimate.  

    Thanks!



    ------------------------------
    Abbie Raikes
    ------------------------------


  • 2.  RE: Trouble getting pooled values

    Posted Fri August 20, 2021 10:29 AM
    Hi. I created a simple dataset with a few missing values, ran MULTIPLE IMPUTATION, then DESCRIPTIVES on the imputed variable, and everything worked correctly, where by correctly I mean I saw this:


    Notice the "Pooled" row at the bottom of the table. Other pooled statistics are available depending on the procedure you are using.
    You don't designate values as imputed; if you look at the MI dataset, you will see that imputed values appear in highlighted cells, for example:



    ------------------------------
    Rick Marcantonio
    Quality Assurance
    IBM
    ------------------------------



  • 3.  RE: Trouble getting pooled values

    Posted Fri August 20, 2021 10:39 AM

    Thanks so much!

     

    For some reason, I don't get the "pooled" estimate at the bottom of the descriptives.  I wonder if this could be because I did not run the multiple imputation in SPSS – it was run in R and then we merged in the imputed values (we had a total of 20 estimates).  So I didn't run the MULTIPLE IMPUTATION command. 

     

    Is there a step I missed? 


    The information in this e-mail may be privileged and confidential, intended only for the use of the addressee(s) above. Any unauthorized use or disclosure of this information is prohibited. If you have received this e-mail by mistake, please delete it and immediately contact the sender.





  • 4.  RE: Trouble getting pooled values

    Posted Fri August 20, 2021 11:07 AM
    Edited by Rick Marcantonio Fri August 20, 2021 11:30 AM
    Try this:

    Be sure that you have a variable named IMPUTATION_, from 0 to the number of imputed datasets.
    SPLIT FILE BY Imputation_.

    then, run DESCRIPTIVES.


  • 5.  RE: Trouble getting pooled values

    Posted Fri August 20, 2021 12:23 PM
    R has a number of imputation packages and methods, so you might or might not be able to match what you used in R with the methods available in SPSS.  And the way that the imputation indicator is saved in R may depend on which package is used.  Imputation using mice, for example, creates a boolean item named where in the imputed output that is separate from the data element that holds the values(data).  AFAIK, the imputation markers are not accessible in SPSS, so you would not have a way to map the R markers into SPSS.

    If performance is an issue for imputation, I gather that you used 20 imputed datasets.  That is an unusually large number to use, so you might consider cutting this down if you are going to do this over again.



    --