Waterfall Plot for an Individual Patient

Usage

plot_patient_waterfall(shap_values, patient_index = 1, top_n = 10)

Arguments

shap_values: The output from explain_kernel().
patient_index: The row index of the patient to explain
top_n: Number of features to show (default 10)

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[1:10, , drop = FALSE],
    X_background = X[11:40, , drop = FALSE],
    nsim = 5
  )

  plot_patient_waterfall(
    shap_values = shap_values,
    patient_index = 1,
    top_n = 5
  )
}