R-squared is 1 minus the residual sum of squares divided by the total sum of squares, so 0.85 means the model accounts for 85 percent of the variation around the mean. It never drops when a term is added, so adjusted R-squared, which penalizes extra terms, and predicted R-squared, which tests the model on left-out points, are better guides. A high R-squared says nothing about whether the model is correct, causal or valid outside the data range.
R-squared is 0.97 but predicted R-squared is 0.41. You're overfitting, drop the three-way terms.
Also heard as
- R2
- coefficient of determination
Related terms
Regression
Fitting an equation that predicts a response from one or more input variables, usually by least squares.
Residual
Difference between an observed value and the value a fitted model predicts for it.
Leverage
Measure of how far a data point's input values sit from the rest, which gives it outsized pull on a fitted model.
ANOVA
Analysis of variance: splitting total variation into parts from each factor and from error, then testing which parts are real.