civic_split: Reproducible train/test split

In civic.icarm: Interpretable Civic-Accountable and Responsible Machine Learning

Reproducible train/test split

Description

Splits a data frame into training and test sets. The seed is always stored in the returned object so the split is fully reproducible. Optional stratification preserves class proportions.

Usage

civic_split(data, prop = 0.75, seed = 2025L, stratify = NULL)

Arguments

data	A 'data.frame' or 'tibble'.
prop	Proportion for training (default '0.75').
seed	Integer random seed (default '2025').
stratify	Optional column name (character) to stratify on. Ensures class proportions are preserved in both splits. Works for both factor (classification) and numeric targets (stratifies by quartile).

Value

A named list with elements 'train', 'test', 'seed', and 'prop'.

Examples

# Any data frame works
splits <- civic_split(iris, prop = 0.8, stratify = "Species")
nrow(splits$train)  # ~120
nrow(splits$test)   # ~30

# Numeric stratification (by quartile)
splits2 <- civic_split(mtcars, prop = 0.75, stratify = "mpg")

civic.icarm documentation built on June 18, 2026, 1:06 a.m.