GEE model with Error " The generalized Hessian matrix is not positive definite"

#1
Hi experts,
When I was trying to build a marginal GEE model (logit = bin) for a censoring variable in a survival study, I implemented the following SAS procedure:
------------------------------------------------------------------------------------------------
proc genmod data = midata5; *uncensoring probablility pr(Cij = 0|Ci(j-1) = 0, time, previous outcome);
class id t;
model y = t prevy/dist =bin link=logit type3 wald scale = pearson;
repeated subject = ID / withinsubject = t type = UN corrw modelse;
run;
------------------------------------------------------------------------------------------------
Here the theoretical model is:
logit[Pr(Cij = 0 | Ci(j-1) = 0, time] = a0 + a1*I (time = 1) + a2*I(time = 2) + a3*I(time = 3) + a4*I(time =4) +a5*Ci(j-1)

However I get the following errors:
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.



And all the standard errors for coefficients are either missing or zero.

Can anyone help me with this problem, please?
Thank you very much!
 

JesperHP

TS Contributor
#2
Could be that the regressor variables are not adding any explanatory value somewhat like when you have multicollinearity in a linear model. Have you tried removing some of the independent variables ?