Programming the statistical procedures from SAS

proc mianalyze on mixed reml

Reply
N/A
Posts: 1

proc mianalyze on mixed reml

Hi,

I'm trying to run a mixed model (repeated analysis) on my imputed database (imputed in spss, then imported to sas). This works fine, however, I would like to pool the estimates of the 5 imputed sets.

When trying to use mianalyze I get an error.

My mixed model:

procmixed method=reml noclprint noitprint data=growth.growth;

  class MeanSys_cat;

  model sdheigth = MeanSys_cat age MeanSys_cat*age AGE_M BMI_0 PARITY2_catd2 EDUCM2_cat_catd1 SMOKE_PRE2d2 FOLIUM_VALIDATED_catd1 FOLIUM_VALIDATED_catd2 SBP_1 DBP_1 hyperdis GENDER agey5child ETHNINFv2_catd2  BFEVER_catd1/solution covb;

  repeated / type=un subject= idm r rcorr;

  by _imputation_ ;

odsoutput SolutionF=mxparms CovB=mxcovb;

  run;

MI analyze:

procmianalyze parms (classvar=full)=mxparms;

  class MeanSys_cat;

  modeleffects Intercept MeanSys_cat age MeanSys_cat*age AGE_M BMI_0 PARITY2_catd2 EDUCM2_cat_catd1 SMOKE_PRE2d2 FOLIUM_VALIDATED_catd1

  FOLIUM_VALIDATED_catd2 SBP_1 DBP_1 hyperdis GENDER agey5child ETHNINFv2_catd2  BFEVER_catd1;

  run;

And this is the error:

ERROR: _Imputation_ less than or equal to 0 in the input PARMS= data set.

Does anyone know what I am doing wrong? Or if it is even possible to do what I'm trying?

Thanks!

Liza

Respected Advisor
Posts: 2,655

Re: proc mianalyze on mixed reml

You have to input the covariance matrix, and that is not easily done.  This link from almost a year ago addresses a lot of my concerns with MIANALYZE and MIXED.

https://communities.sas.com/message/204190#204190

Steve Denham

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