ExactPath.TS: Exact Calculation of Leave-One-Covariate-Out regularization...

In statcao/LOCOpath: R package for high-dimensional inference using LOCO regularization path

Description

This function performs the exact Leave-One-Covariate-Out regularization path estimator

Usage

ExactPath.TS(X, Y, which.covariate, betaNull, multiTest,
  path.method = "lars", norm = "L2.squared", normalize = TRUE,
  intercept = FALSE)

Arguments

X:	a n by p matrix, design matrix
Y:	a n by 1 matrx, response
which.covariate:	a vector or a list of vector, specify which covaritate β_j to perform our test
betaNull:	a vector or a list of vector, should be the same shape as which.covariate. Specify the Null hypothesis, H_0: β_j = betaNull vs H_1 \neq betaNull if using multiTest = FALSE, then will return
multiTest:	boolean, TRUE: test simultaneously, FALSE: test individually
path.method:	string, c("lars", "plus.lasso", "plus.mc+", "plus.scad" )
norm:	string, c("L1", "L2", "L2.squared", "L_inf")
normalize:	boolean, using normalization or not
intercept:	boolean, include intercept or not

Value

a vector of test statistic for each test specified

Examples

X = matrix(rnorm(100*12), 100, 12)
beta = c(1,2, rep(0, 10))
Y = X %*% beta + rnorm(100)
# Test \eqn{H_0: \beta_1 = 1} and \eqn{H_0: \beta_1 = 2} individually 
ExactPath.TS(X, Y, which.covariate = c(1, 2), betaNull = c(1, 2), multiTest = FALSE)
# Test \eqn{H_0: \beta_1 = 1, \beta_1 = 2} simultaneously
ExactPath.TS(X, Y, which.covariate = list(c(1, 2)), betaNull = list(c(1, 2)), multiTest = TRUE)
# Test \eqn{H_0: \beta_1 = 1, \beta_2 = 2} simultaneously and test \eqn{H_0: \beta_1 = 0, \beta_2 = 0}
ExactPath.TS(X, Y, which.covariate = list(c(1, 2), c(1, 2)), betaNull = list(c(1, 2), c(0, 0)), multiTest = TRUE)

statcao/LOCOpath documentation built on July 11, 2020, 6:44 p.m.