Bundles the covariates, the learner, and whether to refit per permutation.
Pass the result as adjust = to riposte_test(), or pass a bare one-sided
covariate formula and let riposte_test() wrap it with the default learner.
Usage
riposte_adjust(covariates, learner = riposte_ridge_learner(), refit = TRUE)Arguments
- covariates
a one-sided formula of covariates, e.g.
~ x1 + x2.- learner
a learner
(X, Y, control) -> Yhat; ridge by default. Seeriposte_ridge_learner().- refit
refit the learner inside every re-randomization (
TRUE, exact for any learner) or fit once on the observed controls and permute the fixed residuals (FALSE, the shortcut that breaks exactness when the learner overfits — kept for comparison).