error with proc mianalyze with multinomial logistic regression

[estul]

Hello all,

Currently trying to get pooled results for my multipley imputed and bootstrapped data for a multinomial regression (see code below), but I get the error  "ERROR: Within-imputation COV matrix is not symmetric for _Imputation_= 1 in the input DATA= data set" when I run my Proc MIANALYZE step. It doesn't seem that my variable names are too long in character so I don't think that's the issue. Any guidance on how to trouble-shoot would be much appreciated.  I have been using the code described on MI for multinomial (nominal) logistic regression here:  http://support.sas.com/kb/49/579.html.

Thanks!

Code:

/*CRUDE MODEL*/
PROC LOGISTIC DATA= final.out_ipw_all outest = final.parms covout descending;
BY _imputation_ replicate;
WHERE sp2=1;
CLASS style(REF='2') income(REF='1') / PARAM=REF;
MODEL style=income/ link = glogit;
WEIGHT ipw_trim;
ODS OUTPUT ParameterEstimates=final.out_crude;
RUN; 

/*proc mianalyze*/
proc mianalyze data = final.parms edf = 550;
modeleffects income2_3 income2_1 income3_3 income3_1 intercept_1 intercept_3;
run;

[SAS_Rob · Accepted solution]

I suspect it has something to do with the BY statement in Proc LOGISTIC.  I noticed that you are not only doing it BY _IMPUTATION_ but also REPLICATE.  You would need a corresponding BY statement in MIANALYZE to take into account that second variable.