I need to use Predictive Mean Matching (PMM) for multiple imputation in SPSS.

According to SPSS support at the following link

https://www.ibm.com/support/knowledgecenter/SSLVMB_24.0.0/spss/mva/syn_multiple_imputation_impute.html

they say "For PMM the imputed value always matches an observed value (specifically, the observed value that is closest to the value drawn by the imputation model).”

SPSS only uses 1 observed value (k = 1) for imputing a missing value. According to Paul Allison, this is insufficient and the default k = 1 should be overridden to k = 5. However, no instructions on how to do this are provided and Allison says in a Q & A at the end of the paper that he does not know how to change the default. This is the link to Allison's article

https://statisticalhorizons.com/predictive-mean-matching

I wrote to SPSS 2 weeks ago but so far haven't heard back. Has anyone come across instructions on how to change the default from k = 1 to k = 5?

Many thanks.

More Judith Fethney's questions See All
Similar questions and discussions