README.md

In crcox/cvblocker: Assign Conditions into Balanced Groups for Cross-Validation

cvblocker

The goal of cvblocker is to provide a simple interface for spliting datasets for k-fold cross validation. The cross validation itself is left up to you, but cvblocker will help split factorial data into balanced groups to iterate over.

Installation

This project is not release on CRAN at the moment.

You can install the development version of cvblocker directly from GitHub with:

install.packages("devtools")
devtools::install_github("crcox/cvblocker")

Example

The factorial input can be coded as numeric, character, or factor.

## basic example code
y.num <- sample.int(3, size = 100, replace = TRUE)
cv <- cvblocker(y.num, k = 6)

y.char <- LETTERS[y.num]
cv <- cvblocker(y.char, k = 6)

y.factor <- factor(y.num, levels = 1:3, labels = c("cow","duck","pig"))
cv <- cvblocker(y.factor, k = 6)

Each time cvblocker runs, it will produce different assignments due to random assignment to groups. Set a random seed for reproducable results. Or, better yet, generate the cross validation blocks one time and save them to a file to be reloaded as needed.

crcox/cvblocker documentation built on May 12, 2019, 5:16 p.m.