Skip to contents

Internal prediction engine: evaluate per posterior draw, then summarize.

dot-predict_mixgpd.Rd

Project rules:

  • density/survival: either provide both (x,y) or neither (defaults to training X and training y).

  • quantile/sample/mean: y must be NULL; x may be provided (new X) or NULL (defaults to training X).

  • CRP predictions use posterior weights derived from z for each draw.

  • Stores per-draw results in object$cache$predict (environment) for reuse in treatment effects.

Usage

.predict_mixgpd(
  object,
  x = NULL,
  y = NULL,
  ps = NULL,
  id = NULL,
  type = c("density", "survival", "quantile", "sample", "mean", "rmean", "median", "fit"),
  p = NULL,
  index = NULL,
  nsim = NULL,
  level = 0.95,
  interval = "credible",
  probs = c(0.025, 0.5, 0.975),
  store_draws = TRUE,
  nsim_mean = 200L,
  cutoff = NULL,
  ndraws_pred = NULL,
  chunk_size = NULL,
  show_progress = TRUE,
  ncores = 1L,
  sample_draw_idx = NULL
)

Details

This is the main internal workhorse behind predict.mixgpd_fit() and the causal effect helpers. It evaluates the requested predictive functional separately for each retained posterior draw, using either explicit SB weights or CRP weights reconstructed from latent cluster labels, and only then collapses the draw-level results into posterior summaries.

The helper also manages caching of per-draw predictive quantities because treatment-effect functions repeatedly reuse the same arm-specific predictive draws. That cache avoids recomputation while keeping the public prediction interface simple.