toys.data.multi: Toys data in multinomial case

In armada: A Statistical Methodology to Select Covariates in High-Dimensional Data under Dependence

Description

toys.data.multi is a simple simulated dataset of a multinomial classification problem.

Usage

toys.data.multi

Format

An object of class list of length 2.

Details

• $Y: output variable: a factor with 3 levels "-1", "0", and "2";

• $x A data-frame containing input variables: with 60 obs. of 50 variables.

The data-frame x is composed by 2 independant clusters, each cluster contains 25 correlated variables. It is an equiprobable three class problem, Y belongs to -1,0,1. There is only 6 true variables, that are in the first cluster, the others being noise. The simulation model is defined through the conditional distribution of the X^j for Y=y. In the first cluster, the X^j are simulated in the following way:

• X^j ~ N(2*y,2) for j=1,2,3,4,5,6;

• the other variables are noise, X^j ~ N(0,1) for j=7,. . . ,25.

The second cluster of 25 variables contains only noise variables.

Examples

library(ClustOfVar)
library(impute)
library(FAMT)
library(VSURF)
library(glmnet)
library(anapuce)
library(qvalue)
X<-toys.data.multi$x
Y<-toys.data.multi$Y
scoreX<-data.frame(c(rep(8,6),rep(0,44)))
rownames(scoreX)<-colnames(X)
select<-ARMADA.heatmap(X, Y,  scoreX, threshold=1)
## Not run: 
result<-ARMADA(X,Y, nclust=2)
select<-ARMADA.heatmap(X, Y,  result[[3]], threshold=5)

## End(Not run)

armada documentation built on May 2, 2019, 6:37 a.m.