Basic regressionBasic regression
- variance explained
- R2 = 100 * (1 - sum(residuals.^2) / sum((y - mean(y)).^2))
- residuals = y - X*(inv(X'*X)*X'*y)
- If there are no residuals, R2 is 100% (1.00).
- If your model simply consisted of a constant term, the R2 of that model is 0% (0.00).
- If you keep adding new predictors, R2 is guaranteed to not decrease.
- additional variance explained
- Fit model M1, get your R2 (A).
- Fit model M2 which subsumes M1 but adds one or more additional predictors, get a new R2 (B).
- B-A is termed amount of additional variance explained