Explain Survival Models via survex and DALEX — explain

Adapts Tidymodels survival models (workflows, parsnip / censored, coxnet_model, and joint_model) to create a survex survival explainer. Standardizes survival predictions into \(N \times K\) survival curves, bridging the gap identified in survex issue #98.

Usage

explain_survival(
  model,
  data = NULL,
  y = NULL,
  predict_survival_function = NULL,
  predict_risk_function = NULL,
  times = NULL,
  label = NULL,
  verbose = FALSE,
  ...
)

Arguments

model: A fitted model object: joint_model, coxnet_model, cv_coxnet, workflow, or model_fit.
data: A data frame containing predictor columns used to calculate explanations. If NULL and model is a joint_model, the training predictors are used, as the original columns (e.g. a factor stays one column, not its one-hot dummies), which is what the model's predict() takes.
y: A survival object or response variable (e.g. Surv(time, status)). If NULL and model is a joint_model, the original survival object is used.
predict_survival_function: Optional custom prediction function returning an \(N \times K\) data frame or matrix of survival probabilities \(S(t)\).
predict_risk_function: Optional custom prediction function returning a numeric vector of risk scores or cumulative hazard.
times: A numeric vector of evaluation time points. If NULL, defaults to unique event times in y.
label: Character string naming the model in plots and summaries.
verbose: Logical; if TRUE, progress messages are shown. Default is FALSE.
...: Additional arguments passed to survex::explain_survival().

Value

A survex_explainer object compatible with survex::model_parts(), survex::model_performance(), and survex::predict_parts().

Examples

if (FALSE) { # \dontrun{
if (requireNamespace("survex", quietly = TRUE) && requireNamespace("survival", quietly = TRUE)) {
  lung <- survival::lung
  df <- na.omit(lung[, c("time", "status", "age", "sex")])
  fit <- joint_model(df, survival::Surv(time, status) ~ age + sex)
  expl <- explain_survival(fit)
  mp <- survex::model_parts(expl)
}
} # }