Hello,
I have fit a shared-frailty model using stcox with the shared() option, and have found that the proportional hazards assumption does not hold for one predictor. To address this, I have attempted to add in the tvc() option for the predictor in question, however, STATA then goes off nowhere (basically, estimation appears to start but does not go any further than "Fitting comparison Cox model ..." and "Estimating frailty variance:" in the output, even overnight). I have not found any information as to whether these two options are mutually exclusive, and judging from the fact that the command is accepted they don't seem to be, but possible something is wrong about the idea? I'd be very grateful if someone could provide hints on how to get around this problem, possible via an alternative command that produces the same model?
Best regards
Hanna
I have fit a shared-frailty model using stcox with the shared() option, and have found that the proportional hazards assumption does not hold for one predictor. To address this, I have attempted to add in the tvc() option for the predictor in question, however, STATA then goes off nowhere (basically, estimation appears to start but does not go any further than "Fitting comparison Cox model ..." and "Estimating frailty variance:" in the output, even overnight). I have not found any information as to whether these two options are mutually exclusive, and judging from the fact that the command is accepted they don't seem to be, but possible something is wrong about the idea? I'd be very grateful if someone could provide hints on how to get around this problem, possible via an alternative command that produces the same model?
Best regards
Hanna
