binarize: Convert categorical variables into binary dummies

In NCI-CGR/construct.model.matrix: Construct Model Matrix for `plco-analysis`

Description

This function performs categorical variable binarization, which is handled in other libraries but in this case is performed with some particularities. Conditions with fewer than a specified number of subjects are combined into a single meta-group tagged "combined.other", which has undesirable statistical properties but at the very least is less likely to cause your downstream model to just implode.

Usage

binarize(df, minimum.factor.level.count, conditions)

Arguments

df	data.frame, source data frame containing model matrix data to be transformed
minimum.factor.level.count	numeric, number of subjects in a condition strictly below which the condition is merged into the catch-all meta group
conditions	list of logical vectors. each vector should be of length ncol(df), and should indicate whether a column passes a particular criterion for being included in the transformed set. each vector is effectively logically ORed together.

NCI-CGR/construct.model.matrix documentation built on Aug. 10, 2021, 8:53 a.m.