Response against one predictor
Observed vs predicted
Residuals vs fitted
Normal Q–Q
Residual distribution
Residuals vs predictor
Bootstrap confidence intervals
Refit the model to 1,000 resamples of the training rows (rows drawn with replacement) and take the 5th and 95th percentiles of each coefficient. Compare with the 90% t-interval from the model formula.
Predictive accuracy on held-out rows
The model never saw these rows. Coverage is the share of test responses inside the 95% prediction interval; a calibrated model lands near 95%.
Test error
Saved models
Save a model to compare it with the others. Best test RMSE is highlighted.
Make a prediction
Set each predictor's value. The star on the fit plot marks the new point.