topic Re: Pooled descriptive statistics after multiple imputation in Statistical Procedures

Pooled descriptive statistics after multiple imputation

UcheOkoro — Mon, 26 Apr 2021 19:38:10 GMT

Please, I need help with SAS code for pooled descriptive statistics after Multiple imputation. I need to calculate the pooled proportions, median with the interquartile range and risk difference. Below is my SAS code, thank you :

proc means data=Avera_matched_data_all n nmiss q1 q3 median;
class telemed_use;
var age;
by _Imputation_;
run;


proc freq data=Avera_matched_data_all;
table  new_dialysis*telemed_use sex2*telemed_use age2*telemed_use/norow nocol chisq riskdiff;
by by _Imputation_;run;

Re: Pooled descriptive statistics after multiple imputation

SteveDenham — Tue, 27 Apr 2021 12:48:23 GMT

PROC MIANALYZE. After doing the by imputation runs, you need to consolidate the results. The first example in the MIANALYZE documentation should provide what you need for the output from PROC MEANS. Unfortunately, PROC FREQ does not provide a standard error for the analysis, so you will need to switch to PROC SURVEYFREQ. See this post from @SAS_Rob for details: https://communities.sas.com/t5/Statistical-Procedures/proc-freq-with-multiple-imputation/td-p/671680

SteveDenham

Re: Pooled descriptive statistics after multiple imputation

UcheOkoro — Tue, 27 Apr 2021 15:23:16 GMT

Thank you so much! It worked.

Re: Pooled descriptive statistics after multiple imputation

UcheOkoro — Tue, 27 Apr 2021 16:25:32 GMT

Please, I have another related question. I am using the proc surveymeans and I would like to include a within stratum median analysis statement

proc surveymeans data=Avera_matched_data_all q1 q3 median;
stratum telemed_use/list;
var age ;
by _imputation_;
run;

I would like something to similar to the following

Re: Pooled descriptive statistics after multiple imputation

SteveDenham — Thu, 29 Apr 2021 13:18:57 GMT

Not totally familiar with this, but based on the documentation, I think you will need to change your STRATA (misspelled, btw) to a DOMAIN statement.

SteveDenham