Simulate cross-fitted predictive data — simulate_crossfit

Generates a synthetic regression/classification dataset with covariates, outcomes, and out-of-fold predictions from a K-fold cross-fitted GLM.

Usage

simulate_crossfit_data(
  n = 20000,
  p = 5,
  family = stats::binomial(),
  K = 5,
  seed = 1
)

Arguments

n: Number of observations to simulate.
p: Number of covariates (columns in the design matrix).
family: GLM family (defaults to stats::binomial(); stats::gaussian() gives a Gaussian response).
K: Number of folds used for cross-fitting.
seed: RNG seed for reproducibility.

Value

A data frame containing the outcome y, covariate columns x1, ..., xp, and fhat_cf, the cross-fitted predictions. Attributes family and K record the generating family and fold count.