Summarize and Visualize C-index Results from Survival Model Cross-Validation

This function aggregates and visualizes the C-index (concordance index) results obtained from cross-validation of multiple survival models. Each model should contain a Resample_matrix element with per-fold C-index values. The function computes the median and MAD (median absolute deviation) of the C-index for each model, identifies the top-performing model, and optionally generates a bar plot summarizing model performance.

Usage

compute_cv_CINDEX(models, file_name = NULL, plot_results = TRUE)

Arguments

models

A named list of survival model objects, where each element corresponds to one fitted model. Each model must contain a Resample_matrix data frame with columns:

c_index

C-index value per resample (numeric).

Resample

Fold or resample identifier (e.g., "Fold1", "Fold2").

file_name

Optional character string used to name the output PDF file saved under "Results/CINDEX_CV_methods_<file_name>.pdf". If NULL, the file is not named explicitly.

plot_results

Logical; if TRUE (default), generates and saves a PDF bar plot showing median C-index ± MAD per model.

Value

A list with the following elements:

CINDEX_summary

A tibble summarizing median and MAD per model.

All_folds

A tibble with raw C-index values from all folds and models.

Top_model

Character string naming the model with the highest median C-index.

Details

For each model:

• The median C-index represents the typical discrimination performance across folds.

• The MAD (Median Absolute Deviation) measures variability in C-index values (robust equivalent of standard deviation).

The plot helps visualize and compare models based on survival prediction performance, with error bars representing ± MAD.

See also

aggregate_results(), predict_and_evaluate_survival()

Examples

if (FALSE) { # \dontrun{
models <- list(
  cox_ph_survival = list(
    Resample_matrix = tibble::tibble(
      c_index = c(0.72, 0.75, 0.70),
      Resample = c("Fold1", "Fold2", "Fold3")
    )
  ),
  rand_forest_aorsf = list(
    Resample_matrix = tibble::tibble(
      c_index = c(0.80, 0.82, 0.78),
      Resample = c("Fold1", "Fold2", "Fold3")
    )
  )
)
compute_cv_CINDEX(models, file_name = "example")
} # }