0

I am currently struggling with the following problem in R:

I would like to use a Cox model with time-varying variables to perform my variable selection via boosting. iCoxBoost takes a Surv object as formula which allows start-stop notation. But when I use it as follows

iCoxBoost(Surv(tstart, tstop, status) ~ ., data=data)

I get the following error message:

Error in '[.default'(response, , "time") : subscript out of bounds

On the other hand coxph(Surv(tstart, tstop, status) ~ ., data=data) as well as iCoxBoost(Surv(time, status) ~ ., data=data) works fine.

Also, when I use GLMBoost I get for

glmboost(Surv(tstart,tstop,status) ~ ., data=data, family=CoxPH(), control=boost_control(mstop = 100))

the following error message:

Error in basefit(u, m) : could not fit any base-learner in boosting iteration 19

On the other hand glmboost(Surv(time,status) ~ ., data=data, family = CoxPH(), control=boost_control(mstop = 100)) works fine.

Are CoxBoost and GLMBoost simply not able to include time-variable variables, or is the problem somewhere else?

benson23
  • 16,369
  • 9
  • 19
  • 38
reginameuser
  • 1

1 Answers1

0

Addendum: I checked with the authors of the R packages; CoxBoost (Version 1.4) and GLMBoost (Version 2.9-7) are indeed currently unable to account for time-varying variables!

Community
  • 1
  • 1
reginameuser
  • 1