ORANGE Navigator & Optimizer Frequently Asked Questions
Question number FAQ-0985
The R² value can sometimes be negative, but since it's the square of the correlation coefficient, shouldn't it be impossible for it to be negative?
Adjusted R² values and cross-estimated R² values can be negative.
*Since a simple squared correlation coefficient can result in R² = 1 even when the model value and the actual value are significantly different, we perform a calculation that ensures R² = 1 only when the difference is 0 (see below for the formula for cross-estimating R²).
Last updated: 2026-07-15