Proc mi analyze problems with pooling

Margret_chidex — Sat, 03 Feb 2024 05:41:46 GMT

Hello everyone.

can anyone please help . I am trying to do a multiple imputation of my missing variables. everything went well until the pooling phase. i have some errors popping up at the pooling phase. the first code I tried is uses the dataset that I generated as output in the analysis phase. But my categorical variables are not in that data set, so stops analysis after reporting that postmeno and the other variables are not there.

I tried another method. i tried combining the dataset mi-brenda and a_brenda together into dataset _all, hoping that would solve my problem. but then I get this error too:

ERROR: Domain error.

ERROR: Termination due to Floating Point Exception

Please can anyone of you help me figure out how to bypass this problem? Thanks.

My codes are displayed below.

/* The Imputation Phase */
 
 
proc mi data= brenda.brenda3 nimpute=10 out=brenda.mi_brenda seed=54321;
   class postmeno k06t1 n01t1 n02t1 r05t1 risk;
   fcs nbiter=10 discrim(postmeno k06t1 n01t1 n02t1 r05t1 risk/details classeffects=include ) reg(eorqlt3 eoralt3
eordft3
eoreft3
eorfst3
eorft3
eorkat3
eorkft3
eornvt3
eorobt3
eorpft3
eorpt3
eorrft3
eorsft3
eorslt3 eorqlt2/details);
   var eorqlt3
eoralt3
eordft3
eoreft3
eorfst3
eorft3
eorkat3
eorkft3
eornvt3
eorobt3
eorpft3
eorpt3
eorrft3
eorsft3
eorslt3 eorqlt2 postmeno k06t1 n01t1 n02t1 r05t1 risk;
run;
 

/*II-The Analysis Phase   */
TITLE " MULTIPLE IMPUTATION REGRESSION ";
proc genmod data = brenda.mi_brenda ;
class postmeno k06t1 n01t1 n02t1 r05t1 risk;
model  eorqlt3= eoralt3
eordft3
eoreft3
eorfst3
eorft3
eorkat3
eorkft3
eornvt3
eorobt3
eorpft3
eorpt3
eorrft3
eorsft3
eorslt3 eorqlt2 postmeno k06t1 n01t1 n02t1 r05t1 risk / link=log dist=normal;
by _imputation_;
ods output ParameterEstimates = brenda.a_brenda;
run;
quit;
 
Data brenda.all;
set brenda.mi_brenda brenda.a_brenda;
run;
 
PROC CONTENTS data=brenda.all;
run;
 
/*III-Pooling phase  */
 
 
proc mianalyze parms(Classvar=Full)= brenda.a_brenda;
class postmeno k06t1 n01t1 n02t1 r05t1 risk;
modeleffects intercept eoralt3
eordft3
eoreft3
eorfst3
eorft3
eorkat3
eorkft3
eornvt3
eorobt3
eorpft3
eorpt3
eorrft3
eorsft3
eorslt3 eorqlt2 postmeno k06t1 n01t1 n02t1 r05t1 risk ;
run;
 
 
proc mianalyze parms(classvar=full)= brenda.alL;
class postmeno k06t1 n01t1 n02t1 r05t1 risk;
modeleffects intercept eoralt3
eordft3
eoreft3
eorfst3
eorft3
eorkat3
eorkft3
eornvt3
eorobt3
eorpft3
eorpt3
eorrft3
eorsft3
eorslt3 eorqlt2 postmeno k06t1 n01t1 n02t1 r05t1 risk ;
run;

Re: Proc mi analyze problems with pooling

SAS_Rob — Mon, 05 Feb 2024 14:08:05 GMT

I think the problem is that you have the wrong sub-option for the CLASSVAR= set. It should be set to LEVEL and not FULL.

proc mianalyze parms(Classvar=level)= brenda.a_brenda;

Re: Proc mi analyze problems with pooling

Margret_chidex — Fri, 09 Feb 2024 16:20:01 GMT

Thank you Rob. I'll try that and get back to you.

Re: Proc mi analyze problems with pooling

Margret_chidex — Fri, 09 Feb 2024 16:25:19 GMT

Thank you Rob. It worked. Im so excited. thank you so very much😊 @SAS_Rob

topic Re: Proc mi analyze problems with pooling in Statistical Procedures

Proc mi analyze problems with pooling

Re: Proc mi analyze problems with pooling

Re: Proc mi analyze problems with pooling

Re: Proc mi analyze problems with pooling