A note on adjusting $R^2$ for using with cross-validation

TL;DR

This paper presents a method to adjust the R^2 statistic for more accurate evaluation of predictive models using leave-one-out cross-validation, addressing potential biases.

Contribution

It introduces a novel adjustment technique for R^2 specifically tailored for cross-validation scenarios, improving model assessment accuracy.

Findings

Adjusted R^2 provides more reliable predictive accuracy estimates.

The method improves model comparison in cross-validation settings.

Applicable to various predictive modeling contexts.

Abstract

We show how to adjust the coefficient of determination ($R^2$) when used for measuring predictive accuracy via leave-one-out cross-validation.