Thanks for the response. It wasn't that APOEn wasn't the variable name, I needed to specify a classvar= option for parm= in the proc mianalyze syntax. I found the more appropriate ods output table name for GEE parameter estimates with model-based standard errors (GEEModPEst), which needed /MODELSE to be added for the repeated statement in proc genmod. I also had to add /dist=bin to the model statement. ods select none; proc genmod data=ce.imputed_data2; class isid depression(ref='0') apoen(ref='0') dementia(ref='0') gender(ref='Män') diabetes(ref='0') cancer(ref='0') heartattack(ref='0') SES(ref='0') educationlevel(ref='0') smokingstatus(ref='0')/param = ref; model depression=apoen dementia gender BMI diabetes cancer heartattack SES Educationlevel smokingstatus/dist=bin covb; repeated subject=isid/MODELSE; by _Imputation_; ods output GEEModPEst=gmparms CovB=gmcovb ParmInfo=gmpinfo; run; ods select all; proc mianalyze parms(classvar=level)=gmparms covb=gmcovb parminfo=gmpinfo; class apoen dementia gender diabetes cancer heartattack SES educationlevel smokingstatus; modeleffects Intercept apoen dementia gender BMI diabetes cancer heartattack SES Educationlevel smokingstatus; run;
... View more
