Programming the statistical procedures from SAS

proc glimmix, abnormally high predicted probabilities

Reply
New Contributor
Posts: 3

proc glimmix, abnormally high predicted probabilities

 

Hello!  

 

I am using proc glimmix (SAS 9.4) for a two level model (health care providers and organizations, which are both categorical) that regresses provider type (0/1) on imaging (0/1).  health care provider is specified as the random intercept.

 

There are 92,100 cases nested in 7,133 providers, nested in 2,370 organizations. All providers and organizations have at least 5 cases and there are no multi-organization providers.

 

The model results are consistent with what I would expect, until I add organization as a fixed effect.  The fixed effect intercept increases considerably, which triples the predicted probability (model 3).  The syntax and models are below.  Any ideas as to what the problem could be?

 

PROC GLIMMIX DATA=please_work METHOD=LAPLACE NOCLPRINT;

CLASS  provider org;

MODEL  X_RAY   (EVENT=LAST) = provider_type org / CL DIST=BINARY LINK=LOGIT SOLUTION ODDSRATIO (DIFF=FIRST LABEL);

random intercept / type=vc subject=provider;

COVTEST / WALD;

run;

 

 

 Model 1model 1bmodel 1cModel 2Model 3
changes to modelnaïve, provider as class, random effects for the provider interceptnaïve, organization as class, random effects for the organization interceptorg and provider as class, org as fixed effect, random provider interceptModel 1A + added provider type to fixed effectsmodel 2 + org as class and fixed effect
Fixed Effects (Solution for Fixed Effects)   
      
Intercept (log odds of receiving a xray with a typical provider)-1.204-1.25810.9022-1.230.5816
 (<.0001)ns(<.0001)(<.0001)NS
type   0.1510.188
    (<.0001)(<.0001)
Odds Ratios     
type   1.161.21
      
Predicted Probability0.2310.2210.7110.2260.641
Error Variance ( Covariance Parameter Estimates)  
Intercept(provider)0.4450.3830.0780.4520.073
 (<.0001)(<.0001)(<.0001)(<.0001)(std error 0)
Model Fit     
(-2) Log Likelihood (lower is better)100,237.00100,847.7095,864.76100,135.0095,729.98
NPI ICC11.91%10.43%2.32%12.08%2.17%

 

Respected Advisor
Posts: 2,989

Re: proc glimmix, abnormally high predicted probabilities

[ Edited ]

Adding a class variable into a model will change the intercept. This is a known fact, based upon the way SAS handles categorical variables. I don't see anything wrong here.

 

Predicted probability of what?

--
Paige Miller
Ask a Question
Discussion stats
  • 1 reply
  • 98 views
  • 0 likes
  • 2 in conversation