filter_gz: Multiple knockoff filter by Gimenez & Zou (2018)

In hugogogo/cheapknockoff: Cost-conscious feature selection using multiple knockoffs

Description

This function implements the stable m-knockoff filter by Gimenez & Zou (2018) given the test statistics (kappa, tau)

Usage

filter_gz(kappa, tau, fdr = 0.2, n_knockoff)

Arguments

kappa	A p vector of test statistics, with kappa_i = 1 indicating the original variable winning
tau	A p vector of test statistics, showing the manitude/importance of the variable
fdr	The pre-specified upper bound level of controlled FDR.
n_knockoff	number of knockoffs constructed

Value

The selected variable and their corresponding scores

Examples

library(cheapknockoff)
set.seed(123)
n <- 100
p <- 30
x <- matrix(data = rnorm(n * p), nrow = n, ncol = p)
y <- x[, 1] - 2 * x[, 2] + rnorm(n)
omega <- c(2, 9, sample(seq(2, 9), size = 28, replace = TRUE))
# construct multiple knockoffs
X_k <- multiple_knockoff_Gaussian(X = x, mu = rep(0, p), Sigma = diag(1, p), omega = omega)
# compute knockoff statistics
stat <- stat_glmnet_coef(X = x, X_k = X_k, y = y, omega = omega)
# run gz filter
gz_output <- filter_gz(kappa = stat$kappa, tau = stat$tau, fdr = 0.2, n_knockoff = max(omega))

hugogogo/cheapknockoff documentation built on Sept. 15, 2021, 1:06 p.m.