CARS: The CARS procedure for controlling the false discovery rate

In CARS: Covariate Assisted Ranking and Screening for Large-Scale Two-Sample Inference

Description

This function runs the CARS procedure, constructing the auxiliary variables, computing the test statistics, choosing the cutoff and selecting the locations.

Usage

CARS(X, Y, alpha, tau = 0.9, variance, option = c("sparse", "regular"))

Arguments

X	the first matrix or data frame of observation
Y	the second matrix or date frame of observation
alpha	targeted FDR (false discovery rate) level
tau	the threshold for choosing interesting locations for density estimation, default is 0.5
variance	for X and Y, default is NULL. If provided, in the form of a m*2 matrix, the columns are representing x and y's variance for each location
option	sparse case for FDR control or regular for FDR control. Sparse case has more stability when it comes to controlling FDR, power is slightly compromised

Value

A list containing the following components:

de	decision for each location (0 or 1)
cars	estimated CARS statistics
th	threshold for CARS procedure

Examples

X <- matrix(rnorm(1000),ncol=5,nrow=200);
Y <- matrix(rep(c(0,3),c(800,200))+rnorm(1000),ncol=5,nrow=200);
CARS(X,Y,0.05,tau=0.9,option='regular');

CARS documentation built on May 2, 2019, 2:57 p.m.