proc mi

superbibi — Tue, 12 Feb 2019 19:58:26 GMT

Hi Friend,

I am new to multiple imputation procedure.

I have a repeated measure data sets (presenting 2 in 200 observation here). Group is not missing for all the participants while baseline, viit2, visit3, or visit 4 can be missing. I plan to use the proc mi code as below.

id	group	baseline	visit2	visit3	visit4
1	1	82.00	79.70	76.60	77.10
2	2	87.70	88.95	84.50	.

proc mi data=a out = a_mi seed=21355417 nimpute=30  ;
   var basline visit2 visit3 visit4 group ;
run;

I am confused whether monotone, mcmc, or fcs statement is necessary for this imputation process?

Thanks.

Re: proc mi

SAS_Rob — Wed, 20 Feb 2019 19:57:27 GMT

If group is a categorical variable (which it sounds like it is) then you should use either the FCS or MONOTONE (and add it to the CLASS statement) statement, depending on the missing data pattern. This would allow you to use the REGRESSION method which would treat group as a CLASS variable.

This section in the documentation is helpful as well.

https://go.documentation.sas.com/?docsetId=statug&docsetTarget=statug_mi_details05.htm&docsetVersion=14.3&locale=en

topic Re: proc mi in Statistical Procedures

proc mi

Re: proc mi