marscr: Perform the marginal screening prior to penalized regression

In yliu433/scZINB: Graphical model estimation for zero-inflated count data

Description

This function removes the covariates that are not even weakly associated with the response variable prior to penalized regression to improve the performance of the variable selection and reduce the runtime.

Usage

marscr(data, thres = 0.1)

Arguments

data	the sample data
thres	either a number between 0 and 1 or an positive integer. If it is a number between 0 and 1, the function selects the variables whose correlation with the response are greater than the number; if it is an integer, the function selects the top thres variables. Default is 0.1.

Value

A square matrix. The ith row provides the most correlated variables with the ith variable.

Examples

data <- genER(10, 0.1, 20, 0.5)$X
gEstM <- marscr(data, thres = 5)

yliu433/scZINB documentation built on Nov. 30, 2020, 9:07 p.m.