Multiple Imputation in GLIMMIX with continuous dependent

mmovahed — Fri, 16 Oct 2020 20:30:24 GMT

Hi SAS members

I am trying to pool my effect estimates across my imputed datasets running a linear regression using GLIMMIX. My dependent variable is a continuous variable and my independent variable is a binary treatment variable adjusting for PS decile.

Here is my current code:

PROC GLIMMIX DATA=moh20crt.ps1_d_tt_response
METHOD=LAPLACE NOCLPRINT;
class biol (ref='TOFA') time patid ;
MODEL
/*radi_lda*/
/*radi_mda*/
radi=biol decile / CL
SOLUTION ;
RANDOM INTERCEPT /subject=patid ;
by _imputation_;
COVTEST / WALD;where month<=8;
ods output ParameterEstimates=a_mvn ;
run;

/* this part works well*/

Then I try to combine the estimate across 20 MI dataset as below

PROC MIANALYZE data=a_mvn ;
MODELEFFECTS biol ;
stderr biol;
RUN;

/* this part does not work*/

I am getting message below in the log

"ERROR: Within-imputation Estimate missing for variable biol in _Imputation_= 1 in the input DATA=data set"

Any thought to resolve the problem

Thanks

Moh

Re: Multiple Imputation in GLIMMIX with continuous dependent

SAS_Rob — Mon, 19 Oct 2020 11:59:53 GMT

You have to use both the CLASS statement and PARMS= input in Proc MIANALYZE to combine the Parameter Estimates from GLIMMIX.

Try proc mianalyze parms(classvar=full)=a_mvn;

class biol;

modeleffects intercept biol decile;

run;

Re: Multiple Imputation in GLIMMIX with continuous dependent

mmovahed — Mon, 19 Oct 2020 14:08:15 GMT

Dear Rob

Thanks for response and solution. It did work!!

topic Re: Multiple Imputation in GLIMMIX with continuous dependent in Statistical Procedures

Multiple Imputation in GLIMMIX with continuous dependent

Re: Multiple Imputation in GLIMMIX with continuous dependent

Re: Multiple Imputation in GLIMMIX with continuous dependent