Vanilla hunting

Returns the hunted signal h as a function

Usage

hunt_vanilla(
  fit_alt_method,
  resids,
  X,
  X.cols = 1:ncol(X),
  trim.outlier = TRUE,
  arg.fit_alt_method = NULL,
  predict_fun_alt = stats::predict
)

Arguments

fit_alt_method

Function with signature fit_alt_method(y, X, ...) that returns a fitted alternative model \(\hat{g} \in \mathcal{G}\). For a fitted g, it must support predict_fun_alt(g, X) for evaluation.

resids

Residuals (i.e., negative scores) of length n from the null model.

X

Covariates of dim n x p.

X.cols

Subset of covariates to hunt. (Default: 1:ncol(X))

trim.outlier

If TRUE, outliers in \(\hat{h}(X)\) will be trimmed from the hunted \(\hat{h}\) using Tukey's IQR rule.

arg.fit_alt_method

Named list of additional arguments passed to fit_alt_method (default to NULL).

predict_fun_alt

Function with signature predict_fun_alt(fit, X) returning a numeric vector of predictions from the alternative-model fit. Default stats::predict.

Value

A function h of signature h(X).