Within imputation observation Error with Proc Glimmix and Miaanlyze

lyton80 — Fri, 13 Oct 2017 02:25:43 GMT

I have a code snippet shown below which is embedded in a model building macro:
....
proc glimmix data=&indata;
class study;
model &outcome = &covariates / s oddsratio covb;
random intercept / subject=study;
by _Imputation_;
ods select ParameterEstimates CovB FitStatistics;
ods output ParameterEstimates=lgparms CovB=lgcovb FitStatistics=FitStats;
run;
.....
my response/outcome variable is binary so I am inclined to specifiy a binary distribution and a logit link to keep the code consistent to my modeling expectations, However a change to this:

proc glimmix data=&indata;
class study;
model &outcome = &covariates / dist=bin s oddsratio covb;
random intercept / subject=study;
by _Imputation_;
ods select ParameterEstimates CovB FitStatistics;
ods output ParameterEstimates=lgparms CovB=lgcovb FitStatistics=FitStats;
run;

gives me this error:
ERROR: Within-imputation observations not found for _Imputation_= 6 in the input PARMS= data set.

Can anyone decode this error?

Re: Within imputation observation Error with Proc Glimmix and Miaanlyze

lyton80 — Fri, 13 Oct 2017 17:56:32 GMT

On close examination, proc glimmix is not producing ParameterEstimates because some imputed datasets are not converging. This is causing the proc mianalyze to throw the error. ("Within-imputation observations not found for _Imputation_= .. in the input PARMS= data set")

Is there any way to skip over non-converging datasets?

topic Within imputation observation Error with Proc Glimmix and Miaanlyze in Statistical Procedures

Within imputation observation Error with Proc Glimmix and Miaanlyze

Re: Within imputation observation Error with Proc Glimmix and Miaanlyze