Plot SHAP Dependence for SuperSurv

Usage

plot_dependence(shap_values, data, feature_name, title = NULL)

Arguments

shap_values: The output from explain_kernel().
data: The original covariate data used for the explanation (X_explain)
feature_name: String name of the column to plot
title: Optional custom title.

Value

A ggplot object visualizing the SHAP values.

Examples

if (requireNamespace("fastshap", quietly = TRUE) &&
    requireNamespace("glmnet", quietly = TRUE)) {
  data("metabric", package = "SuperSurv")
  dat <- metabric[1:80, ]
  x_cols <- grep("^x", names(dat))[1:5]
  X <- dat[, x_cols, drop = FALSE]
  new.times <- seq(20, 120, by = 20)

  fit <- SuperSurv(
    time = dat$duration,
    event = dat$event,
    X = X,
    newdata = X,
    new.times = new.times,
    event.library = c("surv.coxph", "surv.ridge"),
    cens.library = c("surv.coxph"),
    control = list(saveFitLibrary = TRUE)
  )

  shap_values <- explain_kernel(
    model = fit,
    X_explain = X[1:20, , drop = FALSE],
    X_background = X[21:50, , drop = FALSE],
    nsim = 5
  )

  plot_dependence(
    shap_values = shap_values,
    data = X[1:20, , drop = FALSE],
    feature_name = colnames(X)[1]
  )
}