MDR.high.forward: MDR based selection methods for higher-order interacations

In CateSelection: Categorical Variable Selection Methods

Description

MDR based three-stage selection methods for higher-order interacations

Usage

MDR.high.forward(x, y, order = NULL, trace = NULL, alpha = NULL, beta = NULL, 
pvalue = NULL, r2 = NULL, ...)

Arguments

x	A matrix of genotypic data/genetic markers (predictor variables), where the rows are the samples and the columns are the predictors.
y	A numeric vector of phenotypic data (response variable).
order	The order of interaction. Default is 3.
trace	Show computations? Default FALSE.
alpha	Cutoff value for the difference (D1) of coefficient of determination between single modles with and without MRD interactions in the first stage. Default is 0.1.
beta	Cutoff value for the difference (D2) of coefficient of determination between single modle with p interactions and single model with (p-1) interactions in the second stage. Default is 0.05.
pvalue	Cutoff value for p-value in the third stage. Default is 0.01.
r2	Cutoff value for the difference of coefficient of determination in the third stage. Default is 0.02.
...	Other arguments for future methods.

Value

It returns a matrix with the index of selected interactive predictors, and the corresponding adjusted coefficient of determination.

References

Yi Xu, Jixiang Wu, Detecting higher-order interactions of SNP markers associated with three barley agronomic traits (unpublished).

Examples

data(data2)
y <- data2[,1]
x <- data2[,-1]
res <- MDR.high.forward(x,y,order=3)
res

CateSelection documentation built on May 1, 2019, 10:11 p.m.