topic Proc Qlim and Proc mianalyze in Statistical Procedures

Proc Qlim and Proc mianalyze

Shine — Mon, 07 Aug 2017 20:01:11 GMT

Hi, there,

Anyone knows how to solve this problem when I run proc glimmix imputation model with proc mianalyze.

WARNING: Within-imputation StdErr is zero for variable in _Imputation_= 1 in the input PARMS= data set.

Thank you!

Re: Proc Qlim and Proc mianalyze

SAS_Rob — Tue, 08 Aug 2017 16:44:41 GMT

Can you show us the GLIMMIX and MIANALYZE code you used?

Re: Proc Qlim and Proc mianalyze

Shine — Tue, 08 Aug 2017 21:05:01 GMT

HI, Rob,
I followed exactly your suggestion

proc qlim data=scf;

by _imputation_;

model HS = age agesq female /discrete;

model shares = age agesq female/select(y1=1);

ods output parameterestimates=parms;

run;

data parms;
set parms;
parameter=translate(parameter,'_','.');
run;
proc mianalyze parms=parms;
modeleffects HS_Intercept HS_ age HS_ agesq
HS_female shares_Intercept shares_age shares_agesq shares_female _Rho
_Sigma_shares;

run;

It turned out to be _Rho has a missing standard error and SAS stopped
processing.

Any advice is appreciated.

Re: Proc Qlim and Proc mianalyze

SAS_Rob — Thu, 10 Aug 2017 15:22:01 GMT

With no standard error for Rho you won't be able to get combined estimates. I would remove it from the MODELEFFECTS statement to get MIANALYZE to run.

I should mention that the missing standard error might indicate an issue with the models in QLIM themselves. I would check those to make sure there is convergence, its not overspecified in the 2nd model, etc.