Greedy procedure to determine the selection covariate

Source: R/update_selection_step.R

Calculates the gradient of the smoothed check-loss for the residuals and selects the column of X which has the highest absolute correlation.

update_selection_step(X, residuals, tau, h = 0.1, kernel = "Gaussian")

Arguments

X

(matrix)
A matrix of covariates.

residuals

(numeric())
A vector of residuals.

tau

(numeric(1L))
A quantile.

h

(numeric(1L))
The bandwith for smoothing.

kernel

(character(1L))
The kernel for smoothing.

Value

A list with components

  • sel_cov The index of the selected covariates.

  • cor The corresponding correlation.