PROC MI: MNAR - can we specify 2 modelobs option, like 'modelobs=(trtc='Placebo') modelobs=NCMV(k=2)

HuangR89 — Wed, 27 May 2020 06:46:12 GMT

Hello,

I want to run a MI under MNAR framework using monotone regression model. as for the MNAR, I want to impute all the missing values based on observations from TRT='3' group using the NCMV method with K=2.

Attached is my test code. However, it cannot work out, with ERROR ' if the sepcified identification variable W12 in the MODELOBS= option is in the VAR list, it must be completed observed'.

Can anyone help to explain this error or sovle this error? thanks.

Does it work to specify 2 modelobs options like this?

Re: PROC MI: MNAR - can we specify 2 modelobs option, like 'modelobs=(trtc='Placebo') modelobs=NCMV(

SAS_Rob — Mon, 31 Aug 2020 14:19:30 GMT

If you are using a variable as an indentification variable then it cannot also be one of the variables you are imputing. If you remove w12 from both the VAR and MNAR statements then it should run without issue.

topic Re: PROC MI: MNAR - can we specify 2 modelobs option, like 'modelobs=(trtc='Placebo') modelobs=NCMV( in Statistical Procedures

PROC MI: MNAR - can we specify 2 modelobs option, like 'modelobs=(trtc='Placebo') modelobs=NCMV(k=2)

Re: PROC MI: MNAR - can we specify 2 modelobs option, like 'modelobs=(trtc='Placebo') modelobs=NCMV(