get_risk_input: Create input list for 'get_risk'

In benkeser/cvma: Cross-validation-based maximal association

Description

get_risk computes the cross-validated risk of the entire procedure by calling y_weight_control$cv_risk_fn with this input list.

Usage

get_risk_input(split_Y, Ynames, all_fits, V, learners, all_sl,
  all_fit_tasks, all_weight, sl_control)

Arguments

split_Y	The outcome matrix split by relevant validation folds.
Ynames	The names of the outcomes. Used to search all_fits and all_sl.
all_fits	List of all learner fits.
V	Number of folds.
learners	Vector of super learner vectors.
all_sl	List of all super learner weight fits.
all_fit_tasks	List of all learner fit tasks (faster to search over than search over all_fits).
all_weight	List of all outcome weight fits.
sl_control	List of super learner controls.

Value

List with each entry a list with entries: valid_fold (the number of the corresponding fold), Y (outcome matrix in this outer-most validation fold), pred (matrix of outcome predictions for this validation fold from super learner fit in V-1 training folds), and y_weight (vector of outcome weights computed by minimizing V-2 cross-validated risk of composite super learner). get_risk_input is only used to compute the cross-validated risk of composite super learner on the composite outcome and so is only used in the outer most cross-validation layer.

benkeser/cvma documentation built on May 5, 2019, 1:37 p.m.