Using the Knockoff Filter with a Fixed Design Matrix

In MFKnockoffs: Model-Free Knockoff Filter for Controlled Variable Selection

The MFKnockoffs package can also be used to perform controlled variable selection with a fixed design matrix, assuming a linear regression model for the response. In this sense, MFKnockoffs is a superset of the original knockoffs package.

set.seed(1234)

# Problem parameters
n = 1000          # number of observations
p = 300           # number of variables
k = 30            # number of variables with nonzero coefficients
amplitude = 4.5   # signal amplitude (for noise level = 1)

# Generate the variables from a multivariate normal distribution
mu = rep(0,p); Sigma = diag(p)
X = matrix(rnorm(n*p),n)

# Generate the response from a linear model
nonzero = sample(p, k)
beta = amplitude * (1:p %in% nonzero) / sqrt(n)
y.sample = function(X) X %*% beta + rnorm(n)
y = y.sample(X)

First examples

In order to create fixed-design knockoffs, we call MFKnockoffs.filter with the parameter statistic equal to MFKnockoffs.stat.glmnet_lambda_difference. Moreover, since not all statistics are valid with fixed-design knockoffs, we use MFKnockoffs.stat.glmnet_lambda_difference instead of the default one (which is based on cross-validation).

library(MFKnockoffs)
result = MFKnockoffs.filter(X, y, knockoffs = MFKnockoffs.create.fixed, statistic = MFKnockoffs.stat.glmnet_lambda_difference)

We can display the results with

print(result)

The default value for the target false discovery rate is 0.1. In this experiment the false discovery proportion is

fdp = function(selected) sum(beta[selected] == 0) / max(1, length(selected))
fdp(result$selected)

See also

If you want to see some basic usage of the knockoff filter, see the introductory vignette. If you want to look inside the knockoff filter, see the advanced vignette.

MFKnockoffs documentation built on May 2, 2019, 6:33 a.m.