Problems with running PROC GENMOD ERROR: Error in parameter estimate covariance computation.

[bsriv]

Hi all- relatively new to SAS, and I'm running some GEE models (with repeated measures).   Some are successful, but on the following U am having problems:

My log reads:

WARNING: The negative of the Hessian is not positive definite. The convergence is questionable.

WARNING: The procedure is continuing but the validity of the model fit is questionable.

WARNING: The specified model did not converge.

WARNING: Negative of Hessian not positive definite.

WARNING: The generalized Hessian matrix is not positive definite. Iteration will be terminated.

ERROR: Error in parameter estimate covariance computation.

ERROR: Error in estimation routine.

proc genmod data=work.opioid27;
class BJH_MedRec countx (ref="1") racex (ref="1") agegrp (ref="2") narcanx (ref="1") alcoholx (ref="1") moodx (ref="1") sudx (ref="1") yr (ref="2005") suicidex (ref="1") / param=glm;
model psychx=countx racex agegrp narcanx alcoholx moodx sudx suicidex yr / dist=bin link=logit;
repeated subject=BJH_MedRec / type=cs;
lsmeans countx racex agegrp narcanx alcoholx moodx sudx suicidex yr /diff oddsratio cl;
run;

It seems to be an issue with the dependent variable because whenever I remove any predictors the same thing happens.   Thanks!

[Rick_SAS]

Often this issues indicates a model that does not fit the data. The poor fit can be due to many reasons, and it is impossible to say more without seeing your data, Some reasons include collinearity and over-specified models.

See also Steve Denham's answer to a similar question.

[StatDave · Accepted solution]

Unfortunately, convergence can never be guaranteed for any iterative estimation algorithm such as maximum likelihood or Generalized Estimating Equations. Nonconvergence can happen for many reasons, but just as with ordinary logistic models, sparseness is very often the cause of such problems since it can result in some model parameters actually being infinite. Generally, as model complexity increases and sample size decreases, the problem becomes more sparse and more likely to result in convergence problems. Starting with a simple model (that is, one with fewer parameters to estimate) and only adding variables as they can be supported is often a good strategy.