The application demonstrates how Multiple Linear Regression (MLR) works for selected variables from the People dataset. You can choose two predictors (X1, X2) and one response (y). Then you can see the regression plane, main plots for MLR results, play with autoscaling parameters and learn how cross-validation is performed using step-by-step procedure (every step will be shown with a couple of seconds delay, so you can see all details). The following cross-validation options are available: full — full (leave one out), rand 8 seg and rand 4 seg — random splits with 8 and 4 segments, ven 8 seg and ven 4 seg — venetian blinds splits with 8 and 4 segments. Try these options in table view for better understanding.
The MLR model computed in this app does not have an intercept term (b0), therefore centering is important for making good predictions. If you do not center the data the predicted y-values will have a bias. The data rows are automatically sorted according to values of selected y-variable.
Blue points on 3D plot show calibration subset, red points show validation subset for each step.
Number of segments:
Number of objects: