Quadrant Map

This chart displays the focal cases and their k nearest neighbors on a scatterplot with the target on the y-axis and a scale feature on the x-axis, paneled by features. Reference lines are drawn for scale variables at the variable means in the training partition. Again, the values for the prototypes shown in the charts are the predicted values of Sales in thousands [sales].

Looking at the chart ofr Engine size [engine_s], newCar and its 3 nearest neighbers are to the left of the vertical reference line, and we get a clear idea of why newCar's predicted sales are so high. One of its nearest neighbors is the Civic, which has very high sales. Since sales figures are right-skew, it might be more reasonable to predict sales based upon the median value of the nearest neighbors rather than the mean. This can be specified on the Neighborstab in the Predictions for Scale Target group.

Next