Programming the statistical procedures from SAS

Help with MIAnalyze procedure after proc logistic

Reply
N/A
Posts: 1

Help with MIAnalyze procedure after proc logistic

I was wondering if anyone could help point me in the right direction with a problem I'm having with the MIAnalyze procedure. I created a multiple imputations data set with 5 imputations and now would like to combine the results using MIAnalyze. I first ran a proc logistic with a class statement (as all my covariates are categorical) and then ran the MIanalyze procedure. I keep getting an error message that "the variable cat_matchage is not in the covb= data set." I think the reason is that the new data set now has broken down cat_matchage into the various levels (i.e. cat_matchage1, cat_matchage2, etc.). Code below:

proc logistic data=reach.matrixmi;

class

infection (PARAM=REF REF='0')

cat_matchage(PARAM=REF REF='1')

mrace(PARAM=REF REF='White Non Hispanic')

meduc (PARAM=REF REF='<High School')

fincome(PARAM=REF REF='<$35,000')

msmokmo(PARAM=REF REF='NA')

momAgeBirth(PARAM=REF REF='<25'); 

model uni_regout(event="1")= infection cat_matchage mrace  fincome meduc msmokmo momagebirth/ covb;

by _imputation_;

format meduc feduc educationnewf.  fincome  mincome incomenewf.;

where reachdata=2;

ods output parameterestimates=MI.infection covB=MI.covbinfection;

run;

proc mianalyze parms=mi.infection1 covb=mi.covbinfection1;

modeleffects intercept cat_matchage mrace meduc fincome msmokmo momAgeBirth;

run;

Any guidance would be much appreciated thanks!

Respected Advisor
Posts: 2,655

Re: Help with MIAnalyze procedure after proc logistic

It seems to me that this situation comes up on a fairly regular basis, and it has to do with the structure of the covb matrix, especially with CLASS variables.  I think an answer was presented, but basically MIANALYZE does not play well with covariance matrices with categorical values.  I think an answer from Tech Support made its way on here about six months ago, but I cannot google it out today.  So, I would recommend opening a ticket with Tech Support, and when a satisfactory result is obtained, please post it back here.

Steve Denham

Ask a Question
Discussion stats
  • 1 reply
  • 248 views
  • 0 likes
  • 2 in conversation