I am getting the following error when using mlr to do resampling on a conditional inference forest:
Error in Hmisc::rcorr.cens(-1 * y, s) :
NA/NaN/Inf in foreign function call (arg 1)
My code is
surv.task <- makeSurvTask(data = bb_imp2, target = c("timeToEvent", "status"))
surv.learner <- makeLearner(cl="surv.cforest", predict.type="response", mtry=5)
rdesc <- makeResampleDesc(method="CV", iters=2)
r = resample(surv.learner, surv.task, rdesc)
r
For testing I have cut the data down to just 3 columns - times, status and one predictor - Glucose - and there are no NA, Inf or NaN values in the data. I can successfully train a cforest model on the data without using resampling:
surv.task <- makeSurvTask(data = bb_imp2, target = c("timeToEvent", "status"))
surv.learner <- makeLearner(cl="surv.cforest", predict.type="response", mtry=5)
blood_cforest <- train(surv.learner, surv.task)
getLearnerModel(blood_cforest)
Random Forest using Conditional Inference Trees
Number of trees: 500
Response: Surv(timeToEvent, status, type = "right")
Input: Glucose
Number of observations: 873
There are no character variables in the data. The structure of the data is:
str(bb_imp2)
'data.frame': 873 obs. of 3 variables:
$ timeToEvent: num 373 2934 397 3005 2930 ...
$ status : int 0 0 0 0 0 0 0 0 0 0 ...
$ Glucose : num 5.2 5.6 6.7 5.5 6.1 5.9 5.8 5.7 7.7 5.5 ...
Why does the resampling not work? Where should I upload the data file so that others can try it out?