PROC MI for Chi-Square

[lady8506]

Hi folks,

I am trying to learn about multiple imputation as a way to fill in missing data for a dataset I have. I have found a great tutorial at https://stats.idre.ucla.edu/sas/seminars/multiple-imputation-in-sas/mi_new_1/ that goes into detail. However, there is a part where I am confused because they use PROC MI and subsequent steps to run Linear Regression and Logistic Regression. I would like to use PROC MI, and then I suppose PROC MIANALYZE, for Chi-Square and Frequency tables (PROC FREQ) as my goal is to calculate frequencies and to see if there is a significant difference between variables and the groups they belong to.

All of my variables are binary (e.g. 0 for no insurance, 1 for insurance; 0 for Age less than 50, 1 for Age greater than 50).

Is it possible to use SAS' multiple imputation to run frequency tables and chi-squared analyses afterwards? If so, how? Unfortunately, the above tutorial doesn't help me figure this out.

[SAS_Rob · Accepted solution]

Yes, you would run the Proc FREQ by _IMPUTATION_ saving the Chi-Square statistics to a SAS data set.  Then you could make use of something like Dr. Paul Allison's COMBCHI macro which will give you a single test statistic and p-value.

http://www.ssc.upenn.edu/~allison/combchi.sas

[Oluwole]

Please, could you explain this further? How to save the chi square and how to use the combchi macro?
Thank you