Discriminant Analysis with imputed datasets

[Smith, Gail]

 

I have a dataset containing variables that have some missing values.  I want to run a discriminant analysis.  I have used the multiple imputation procedure in SPSS and have created 20 datasets.   When I run the DA, I get separate results for each dataset.

 

How do I combine/pool these results ?

 

Thanks for any suggestions,

 

Gail

 

Gail L. Smith

Senior Research Analyst

Boys Town National Research Institute

      for Child and Family Studies

(402)-498-1263

[hidden email]

 

[Bruce Weaver]

SPSS computes pooled estimates for some things.  E.g., see the table of coefficients at the bottom of this page:  http://www.uvm.edu/~dhowell/StatPages/More_Stuff/Missing_Data/MissingDataSPSS.html.

But notice that there is no "Pooled" result in the ANOVA table.  (IIRC, someone posted links to some syntax or macros for generating "pooled" versions of the F-tests.)  

HTH.

Smith, Gail wrote

I have a dataset containing variables that have some missing values.  I want to run a discriminant analysis.  I have used the multiple imputation procedure in SPSS and have created 20 datasets.   When I run the DA, I get separate results for each dataset.

How do I combine/pool these results ?

Thanks for any suggestions,

Gail

Gail L. Smith
Senior Research Analyst
Boys Town National Research Institute
      for Child and Family Studies
(402)-498-1263
[hidden email]