IDI.INF function to compare the performance of two Cox regression models

Question

I would like to use the IDI.INF function from the survIDINRI package to the performance of two Cox regression models. For this purpose, I have used the following script:

> D=subset(Dataset, select=c("time","status","Var1","Var2","Var3", "Var4", "Var5", "Var6", "Var7", "Var8"))
>  D$status=as.numeric(D$status==2)
>  D=D[!is.na(apply(D,1,mean)),] ; dim(D)
[1] 800  10
> head(D)
> outcome=D[,c(1,2)]
> covs1<-as.matrix(D[,c(-1,-2)])
> covs0<-as.matrix(D[,c(-1,-2, -10)])
> head(outcome)
> head(covs0)
> head(covs1)
> x <-IDI.INF(outcome, covs0, covs1, t0, npert=300,npert.rand = NULL, seed1 = NULL, alpha = 0.05)

At the end, I get the following message:

Error in fitter(X, Y, strats, offset, init, control, weights = weights,  : 
  NA/NaN/Inf in foreign function call (arg 6)
In addition: Warning message:
In fitter(X, Y, strats, offset, init, control, weights = weights,  :
  Ran out of iterations and did not converge

Are there people who have experience with this IDI.INF function? Are there ways to tackle this problem?

This is difficult for us to work out without your data, or data that exhibits the same problem. — Spacedman, Mar 04 '18 at 13:55
I believe the package documentation says you are supposed to supply only complete cases. — IRTFM, Feb 06 '19 at 04:49

score 0 · Answer 1 · answered Dec 19 '18 at 08:47

This error message could be encountered because of weak predictive value of one variable (might be 6th variable in data.frame). You can fix this problem by removing that variable or including more number of study subjects if possible to make the prediction models stable.

IDI.INF function to compare the performance of two Cox regression models

1 Answers1