CrossValidate (version 2.3.5)

balancedSplit: Split a dataset into training and testing sets, balancing a factor

Description

When performing cross-validation on a dataset, it often becomes necessary to split the data into training and test sets that are balanced for a factor. This function implements such a balanced split.

Usage

balancedSplit(fac, size)

Value

Returns a logical vector with length equal to the length of fac. TRUE values designate samples selected for the training set.

Arguments

fac

A factor that should be balanced between the two subsets.

size

A number between 0 and 1 indicating the fraction of the dataset to be used for training.

Author

Kevin R. Coombes <[e-mail]>

Details

This function randomly samples the same fraction of items from each level of a factor to include in a training set. In most cases, this will be a binary factor (and might even be the outcome that one wants to predict). However, the implementation works for factors with an arbitrary number of levels.

See Also

CrossValidate, CrossValidate-class.

Examples

nFeatures <- 40
nSamples <- 2*10
dataset <- matrix(rnorm(nSamples*nFeatures), ncol=nSamples)
groups <- factor(rep(c("A", "B"), each=10))
balancedSplit(dataset, groups)