I am trying to fit a survival analysis in R with non-recurrent events and time-varying coefficients. The baseline distribution is exponential or Weibull and the frailty distribution is gamma distributed. I have roughly 900.000 rows.

So far I have tried the parfm and frailtypack. Though, neither has worked – they just keep running and never return. The calls for parfm and frailtypack are similar to respectively:

             data= regressionData, hazard="Weibull", RandDist=”Gamma”)

      regressionData, dist="exponential", frailty = "gamma")

Where event is zero-one coded. My guess so far is to use the lme4 package with the function glmer where the family is Poisson, the respond are zero-one coded, the offset is the difference in time and random effect is an intercept for the year factor. I.e. something like:

glmer(event ~.-start-stop-event-year+(1|year), family = Poisson(), offset=stop-start)

I know that this will yield Gaussian distributed random effects and not Gamma. Further, I am not sure that I get the model I want. My goal is to have exponential distributed conditional waiting times and hence I chose the Poisson distribution. Question is whether this is correct? Any suggestions on other packages that will do the job?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.