topic Re: Glimmix model resulting in "blank" standard error in Statistical Procedures

Glimmix model resulting in "blank" standard error

Tissue5454 — Mon, 29 May 2017 02:02:28 GMT

Hi all!

I run logistic regression model with random effiect for binary data by GLIMMIX.

Convergence criterion satisfied and all independent value shows "Estimate"

in Table "Solution for fixed Effects" and "Type III Tests of Fixed Effects".

However, some independent value show blank(".") standard error in

"Solution for fixed Effects" and also in "Type III Tests of Fixed Effects".

Anyone know what kind of situation show this result?

Effect	Estimate	Standard Error	DF	t value	Pr>\|t\|

Intercept	1.6624	0.3209	Infty	5.18	<.0001
D1009	-0.03593	0.007111	Infty	-5.05	<.0001
D1011	-0.00108	0.007111	Infty	-2.67	0.0076
D1012	0.000593	.	.	.	.
D1015	0.00117	.	.	.	.

--------------------------------------------------------------------------------

proc glimmix data=logitbank method=quad;
class fid;
model ydat2(event='1') = D1009 D1011 D1012 D1015 D1019 D1020 D1026 D1031 D1033

D1034 D1035 D1036 D1037 D1039 D1040 D1041 D1045 D1047

D1048 D1050 D1054 D1055 D2002 D2003 D2005 D2006 D2008

/dist = binary link=logit ddfm=none solution;
random intercept/ subject=fid;
output out =glmout pred =xbeta pred(ilink) =predprob;
covtest;
run;

---------------------------------------------------------------------------------

Thanks,

Tissuee5454

Re: Glimmix model resulting in "blank" standard error

sld — Mon, 29 May 2017 19:00:37 GMT

Your model is a multiple regression with 27 predictor variables.

I would look at

(1) whether there is enough data to to support the estimation of 27 parameters; for example, if your sample size is 20, then this model is overparameterized (having "Infty" df suggests overparameterization), and

(2) whether there is multicollinearity among the 27 predictors

(3) whether you have the correct model for your study design

Re: Glimmix model resulting in "blank" standard error

Tissue5454 — Tue, 30 May 2017 09:23:44 GMT

Hi sld!

Thank you for your reply!

There is enough data (n=400,000)

but I have thought of something about multicollinearity.

I should think about it.

Thank you for your advice!

Tissuee5454