MAIL_Full_Stable: MAIL_Full_Stable

In hwyneken/MAILPackage: Model-Averaged Inferential Learning

MAIL_Full_Stable

Description

MAIL_Full_Stable runs MAIL without data splitting, and uses a more stable variable selection method

Usage

MAIL_Full_Stable(XMat, yVec)

Arguments

XMat	a n by p numeric matrix
yVec	a n by 1 numeric vector

Details

This is a specific use of the MAIL function. The function uses the following arguments with MAIL:

• splitOption = "Full"

• firstSOILWeightType = "BIC"

• smallestModelWeightType = "AIC"

• firstSOILPsi = 0.5

• smallestModelPsi = 0

• numSelectionIter = 100

• sigma2EstFunc = "Nested_LPM_AIC_CV_50Split"

• verbose = FALSE

Value

resList a list with the following elements:

• selectedSet: The indices of the selected variables (out of 1,...,p).

• tempCI: A matrix (numSelected by 2) - each row corresponds to the 95

• margVar: A vector (length numSelected) of the sampling variances.

• betaHat: The vector (length p) of model-averaged estimates. Unselected variables have an estimated beta of zero.

• estSigma2: The estimated residual variance.

• candMat: An indicator matrix of candidate models (numModels by p).

• origSelectedSet: The variables selected the initial sweep by SOIL.

See Also

MAIL and MAIL_Split

Examples

two main ideas to explain with an example
 
 how to run MAIL_Full

 how to run diagnostics for MAIL_Full - is it a reliable method for the current data


#### how to run MAIL_Full

# first step - organize the data
# gene expression problem: response is the expression of gene probe X249.
# install package colonCA through Bioconductor (use these steps)
# https://bioconductor.org/packages/release/data/experiment/html/colonCA.html
require(colonCA)
data(colonCA)
dataSet <- t(colonCA@assayData$exprs)
colon_y <- dataSet[,249]
colon_x <- dataSet[,-1*249]
colon_x <- t(unique(t(colon_x))) # remove duplicate columns - original has p = 1999, reduce to p = 1990
colon_y <- scale(colon_y)
colon_x <- scale(colon_x)

colonMAILFull <- MAIL_Full_Stable(XMat = colon_x,yVec = colon_y)

# compare the confidence intervals for MAIL_Full and MAIL_Split
# for X257

hwyneken/MAILPackage documentation built on July 27, 2022, 12:46 a.m.