Error in ranger_surv.unify when type= "chf" #38

Activity

1. kapsner commented on Aug 1, 2024

   

   kapsner

   on Aug 1, 2024

   Contributor

   More actions

   Hi @GLRXV ,

   I am not able to reproduce the error with threeshap v0.3.1:

   library(treeshap)
   
   data_colon <- data.table::data.table(survival::colon)
   data_colon <- na.omit(data_colon[get("etype") == 2, ])
   surv_cols <- c("status", "time", "rx")
   
   feature_cols <- colnames(data_colon)[3:(ncol(data_colon) - 1)]
   
   x <- model.matrix(
     ~ -1 + .,
     data_colon[, .SD, .SDcols = setdiff(feature_cols, surv_cols[1:2])]
   )
   y <- survival::Surv(
     event = (data_colon[, get("status")] |>
                as.character() |>
                as.integer()),
     time = data_colon[, get("time")],
     type = "right"
   )
   
   set.seed(123)
   ranger_num_model <- ranger::ranger(
     x = x,
     y = y,
     data = data_colon,
     max.depth = 10,
     num.trees = 10
   )
   
   
   
   # to save some time for these tests, compute model here once:
   unified_model <- ranger_surv.unify(ranger_num_model, x)
   
   unified_model2 <- ranger_surv.unify(ranger_num_model, x, times = c(23), type = "chf")

   Could you please try, if this is working for you as well?

   👍React with 👍1Reacted by Michael Mayer
2. mayer79 commented on Apr 24, 2026

   

   mayer79

   on Apr 24, 2026

   Contributor

   More actions

   Can't reproduce as well.
3. mayer79

   closed this as completedon Apr 24, 2026