06-09-2015 07:48 PM
I'm trying to fit a Bayesian Piecewise Exponential model using PROC PHREG. But, I'm getting the error below. Can someone please guide me to solve this error.
My code and details of the data are at the end. Let me know if you need more info. about the data. Thanks.w
NOTE: Convergence criterion (GCONV=1E-8) satisfied.
ERROR: Floating Point Zero Divide.
ERROR: Termination due to Floating Point Exception
NOTE: The SAS System stopped processing this step because of errors.
PROC PHREG data=lib.sample;
class X1 X2 X3 ;
model TIME_mon*status(0)=X1 X2 X3 X4 X5 X6 ;
bayes nbi=1000 nmc=10000 thinning=5 seed=1980 plots=all
coeffprior= uniform piecewise=hazard (ninterval=4);
X1, X2, X3 are categorical variables and X4, X5, X6 are continuous variables. TIME_mon is survival time in months (discrete). It's right censored data.
06-10-2015 08:11 AM
This is really flailing in the dark, but do you have any "pathological" data points? Something in one of the classes that doesn't fit piecewise could do this. The piecewise function doesn't have to be differentiable at the join point, but it does have to be continuous. If for some level of one of the categorical variables, the continuous variable shifts dramatically, this collinearity could cause this jump. It sounds almost like a quasi-separation problem in logistic regression, for some reason.