proc genmod regression returns coefficients but no stnd errors or p-values

RobertWF1 — Tue, 15 Aug 2023 21:05:32 GMT

I'm running a repeated measures multiple regression for a healthcare dataset. Each member has two records, one pre-intervention and one post-intervention outcome value (# of skilled nursing facility visits).

My code looks something like this:

ods output GEEEmpPEst=parms;
proc genmod data=health_data;
   class member_id;
   y = time*class time class x1 x2 x3 / dist=negbin ;
   repeated subject=member_id / type=ind;
run;
ods output close;

When I run the procedure the regression coefficients are calculated but the stnd errors, confidence intervals, and p-values are blank.

Any ideas what's going wrong?

                            Algorithm converged.                                                       


                                                       GEE Model Information

                                    Correlation Structure                           Independent
                                    Subject Effect                  hce_member_id (1199 levels)
                                    Number of Clusters                                     1199
                                    Correlation Matrix Dimension                              8
                                    Maximum Cluster Size                                      8
                                    Minimum Cluster Size                                      2


                            ERROR: Error in estimation routine.                                        


                                                         GEE Fit Criteria

                                                       QIC            0.0000
                                                       QICu           0.0000


                                                Analysis Of GEE Parameter Estimates
                                                 Empirical Standard Error Estimates
 
                                                           Standard   95% Confidence
                              Parameter           Estimate    Error       Limits            Z Pr > |Z|

                              Intercept           223254.9    .        .        .         .      .    
                              time*class          -48.2170    .        .        .         .      .    
                              time                -130.091    .        .        .         .      .    
                              class               150.1242    .        .        .         .      .    
                              x1                  -120.939    .        .        .         .      .    
                              x2                  -11.0064    .        .        .         .      .    
                              x3                   -5.7767    .        .        .         .      .

Re: proc genmod regression returns coefficients but no stnd errors or p-values

ballardw — Wed, 16 Aug 2023 08:51:17 GMT

Your first clue is likely the text "ERROR: Error in the estimation routine.

You haven't provided any example data but I suspect this is going to come down to something in, or not in, the data.

Have you tried reducing the data set to maybe 10 or 15 Member_ids and see if that provides more of what you expect? If not, you may want to share the complete log from the trial. That way someone that knows more about Genmod than I do may see something that provides clues.

topic proc genmod regression returns coefficients but no stnd errors or p-values in Statistical Procedures

proc genmod regression returns coefficients but no stnd errors or p-values

Re: proc genmod regression returns coefficients but no stnd errors or p-values