splitTestTrain_resampling: Assign train/test labels over several resamplings of the...

Description Usage Arguments Details Value Examples

View source: R/splitTestTrain_partition.R

Description

Assign train/test labels over several resamplings of the data.

Usage

1
splitTestTrain_resampling(pheno_DF, nFold = 3L, predClass, verbose = FALSE)

Arguments

pheno_DF

(data.frame) table with patient ID and status. Must contain columns for Patient ID (named 'ID') and class (named 'STATUS'). Status should be a char; value of predictor class should be specified in predClass param; all other values are considered non-predictor class Expects rows with unique IDs Rows with duplicate IDs will be excluded.

nFold

(integer) number of resamplings. Each sample will be a test sample exactly once.

predClass

(char) name of predictor class

verbose

(logical) print messages

Details

This function is useful when feature selection needs to occur over multiple resamplings of the data, as a strategy to reduce overfitting. Each sample serves as a test for exactly one resampilng, and as a training sample for the others. The method is provided with the positive label and splits the samples so that an even number of positive and negative classes are represented in all the resamplings (i.e. it avoids the situation where one resampling has too many positives and another has too few).

Value

(list) of length nFold, each with char vector of length nrow(pheno_DF). Values of 'TRAIN' or 'TEST'

Examples

1
2
data(pheno) 
x <- splitTestTrain_resampling(pheno,predClass='LumA')

netDx documentation built on Dec. 11, 2020, 2:01 a.m.