A Couple of Comments About Ch. 4 Material Covered Week 5
Be sure to read about the confounding of the student and balance variables in the discussion of the example considered on pp. 137-139. It's a real good example of a commonly occurring phenomenon.
On both the left and right of FIGURE 4.9 on p. 154, not choosing the best method isn't nearly as harmful as it could be in cases in which the two classes
overlap more. Particularly, of the left side, both decision boundaries cut through regions in which there are relatively few data points (and so if future cases to be classified are similar to the
training data, the two
classifiers will result in about the same error rate, even though the shape of the boundaries are appreciably different).