0
$\begingroup$

I'm doing a random forest survival analysis for a school project and I'm confused about the C-index output that I'm getting relative to the Requested performance error. Shouldn't my C-index get higher the lower my Requested performance error is? It seems intuitively that it should be around 1 - the Requested performance error, but the scores are the same. Am I misinterpreting this? I'm using the codes listed in this link as a guide: https://www.randomforestsrc.org/articles/survival.html

> # Summary of the model
> rf_surv
                         Sample size: 3055
                    Number of deaths: 2004
                     Number of trees: 500
           Forest terminal node size: 15
       Average no. of terminal nodes: 144.496
No. of variables tried at each split: 4
              Total no. of variables: 11
       Resampling used to grow trees: swor
    Resample size used to grow trees: 1931
                            Analysis: RSF
                              Family: surv
                      Splitting rule: logrank *random*
       Number of random split points: 10
                          (OOB) CRPS: 0.14017751
   (OOB) Requested performance error: 0.2420692

> # Getting the C-Index
> get.cindex(rf_surv$yvar[,1], rf_surv$yvar[,2], rf_surv$predicted.oob)
[1] 0.2420692
$\endgroup$

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.