Statistical Regression and Classification:
from Linear Models to Machine Learning

Errata, Commentary and Typos

• p.xxxiv: "Schmidt" should be "Schmid."
• p.10: The number 190.3596 should be 192.5600, and the interval changed accordingly.
• p.22: In the first sentence, paragraph three, regestpts should be predpts. Also, predpts must be a matrix , even if only one row.
• p.178: There is a color insert following this page. One reader reports that his/her copy of the book has the insert in black-and-white. All other copies known to me are indeed in color, but I have placed the color version here.
• p.281: "...between two..."

• In Section 4.18 on neural networks, note that it is customary to scale the data before calling neuralnet() or analogous function in other packages. This custom did not arise from theoretical considerations, but rather because many users found that the iteration processes failed to converge without scaling. (This rather ad hoc, "folklore" approach is typical of machine learning methods.)

  The types of scaling used seem to vary, ranging from use of the R scale() function to simply linearly transforming data to [0,1] or [-1,1]. In the regression case, the "Y" variable (response) is scaled as well.