Errata
See also the page of updates to Apophenia that may affect readers of the book.
p 46, Q2.8: "Demonstrate this via a program that finds up to three triangular numbers for every number from 1 to 100." Should read: "Demonstrate this via a program that, for every integer between 1 to 100, finds up to three numbers that sum to that integer."
p 190, just over line (B): "a function pointer that takes in a double and returns a pointer to int." Should read: "a function pointer that takes in a double and returns an int." That is, line (B) returns a plain int, not a pointer to int. [1]
p 223, MSE calculation, third line: the middle term of the second line shows the negation of the bias, so the middle term of the third line should be negative, instead of positive as shown. Fortunately, this cross-term evaluates to zero, and disappears by the next line. [2]
p 224, Equation 7.1.3: An expansion of (x+y)2 takes the form x2 + y2 + 2xy; similarly, the final term of the main sum should be two times the expression listed. Again, this cross-term evaluates to zero, and thus does not appear in Equation 7.1.4. [2]
p 233, gsl_vector_sort should be gsl_sort_vector. [3]
p 240, variance of the multinomial distribution: the element in the first column, kth row, is listed as -pkp2, but it should of course be -pkp1. [2]
p 276, first line in the first math display: X'β should be Xβ. [2]
p 311: (including the parameters for the constant term) should be (excluding the parameters for the constant term). Thus, after running a regression on K parameters including the constant term, the contrast matrix that would generate this statistic would be K-1 rows.
p 358: The Beta distribution has variance less than 1/12 for unimodal distributions, and greater than 1/12 for bimodal, as per the two examples.
[1] Thanks to Tyler Smith of McGill University, Canada
[2] Thanks to Damion Milne of the Institute for Cancer Research and Treatment, Italy
[3] Thanks to George Delic of HiPERiSM Consulting, LLC