search

jkropko / coxed

Duration-Based Quantities of Interest and Simulation Methods for the Cox Proportional Hazards Model
23 stars 6 forks source link

Simulating data with TVC and a binary covariate #4

Closed jjharden closed 3 years ago

jjharden commented 3 years ago

From a user:

I am trying to implement your function in the coxed package to simulate some survival data. In particular time-varying effects data. I find the function and the package very useful and straightforward. I however want to use binary covariates and not continuous covariates. This isn’t explicitly described in the vignette, so I tried a little hack. The code is below; after some trial and error, I’m nearly certain the error is because of the binary covariate. Any advice would be greatly appreciated.

pseudo <- data.frame(x=rep(c(0,1),each=50)) x.pseudo <- dplyr::select(pseudo,x) my.hazard <- function(t){0.8(0.5^0.8)t^(-0.2)} beta.mat <- data.frame(beta1=-0.4*log(1:100+1))

simdata <- sim.survdata(T=100, num.data.frames=1, hazard.fun=my.hazard, X=x.pseudo, type="tvbeta", beta=beta.mat, censor = 0.2)

jkropko commented 3 years ago

Found bug in code, solved in coxed version 0.3.5