Competing Risks with Relative Entropy Integration with a Competing Risk Prior Model

This function implements the Relative Entropy (RE) framework for discrete-time competing risk models with the prior model is a competing risk prior model.

Usage

CompRiskRE_CP(
  beta_cor = 0.9,
  eta,
  N_ext = 5000,
  N_loc = 2000,
  N_val = 200,
  N_test = 5000,
  nCause = 2,
  mu = 1,
  sigma = 0.05,
  seed = 123
)

Arguments

beta_cor

Numeric correlation level between external and local models. Must be one of 1.0, 0.9, 0.7, 0.5, 0.4, 0.3, 0.1, 0.

eta

Numeric vector of Relative Entropy regularization parameters.

N_ext

Sample size for external data (default: 5000).

N_loc

Sample size for local data (default: 2000).

N_val

Sample size for validation data (default: 200).

N_test

Sample size for test data (default: 5000).

nCause

Number of competing causes (default: 2).

mu

Mean of covariates (default: 1).

sigma

Standard deviation of covariates (default: 0.05).

seed

Random seed for reproducibility.

Value

A list with two components:

models

Fitted models: prior, local, joint, and KL (per eta value).

metrics

Performance results including validation/test deviance, C-indices, and AIC across eta values.

Examples

if (FALSE) { # \dontrun{
eta <- generate_eta(method = "exponential", n = 30, max_eta = 30)
res <- CompRiskRE_CP(beta_cor = 0.9, eta = eta, seed = 2024)
res$metrics
} # }