As an added complication, a softcover student edition of the book was published in 2021, which has exercises after each chapter. For now we only post errata for the original format (no exercises), since the page numbering is different. Apart from the exercises, the content in the two books are identical.
Errata since corrected March 2021 online version.
Page 81, table 6.3 (James Fiedler and Chuck Stanton): The third row
$\widehat{\mbox{sd}}(t)$ in
the table is incorrect. It was computed using formula (6.21) with
exponent $2$ rather than $2x$.The correct numbers should be 1.09 2.20 3.35 4.58 5.91 7.40 9.14 11.3 14.6 22.4
Page 146, Section 9.5 (Asbjorn Thomsen, Malter Nikolasjen, Leander
Kristenson, Pi Madsen): The E-step we present is not quite
correct. We are required to compute the conditional expectation of
the full log-log likelihood. This will include conditional
expectations of missing coordinates as in (9.43), but also also
conditional expectations of their squares. The current formulation
will underestimate the variance of coordinate 2, and
overestimate the correlation.
Page 154, near middle (Junzhe He): missing $du$ in integral near
middle of page
Page 225, equation 12.70 (Junzhe Hw): last term should have
$\hat{\mu}_i$, not $\hat{\mu}_1$
Page 277, Figure 15.3 (Jens Klenke): Holmes rejects for the smallest 6
p-values. The 7th is the first not to be rejected.
Page 347, equation (17.18) (Brendon Greenwell): Should really include
an unpenalized intercept in this equation, which is what actually
happens when we fit this model
Page 377, equations 19.2 and 19.3 (Daniele De Martini ): $x'\beta$
should be $x_i'\beta$.
Errata since corrected December 2017 online version, and corrected for March 2021 online version.
Page 19, Figure 2.2 (Francisco Fonseca): the values .8, .6, .4 etc
are threshold values for x, not c. The c values are 2.75, 1.75, .75,
..., -3.25
Page 49, line 4 (Ari Pakman): the value for the t-statistic in page
9 was 3.01, not 3.13
Page 67 (Ari Pakman): $\mbox{cov}_{\alpha}\{y\}$ in (5.57) reappears as $\mbox{cov}_{\alpha}(y)$ in (5.59)
Page 95, equation (7.23) (Lei Zhao): The equations should be $\frac{1}{n} \big[ (n+0.75) * ( \mathrm{sin} ( \frac{\hat{\mu}^{JS}_i}{2\sqrt{n+0.5}}) )^2 - 0.375 \big]
$
Page 102, equation 7.43 (TH): the inverse should be a transpose
Page 102, equation (7.46) (Sungil Kim): there should be 1/1000
inside each brace
Page 159, Section 10.2, line 8 (Evens Salies): ... frequentist
standard error ...
Page 194, (11.40) (Joshua Hill): the summand in the denominator should
use $\hat{\theta}_{ ( i ) }$ instead of
$\hat{\theta}_{i}$
Page 222, above 12.56 (Ari Pakman): the reference to (12.51) should be to (12.39)
Page 226, equation (12.74)(Sungil Kim): $\hat{\mu}_i$ should be
$y_i$.
Page 227, first line (Ari Pakman): the reference to (3.16) should be
to (3.17)
Page 242, equation (13.28) (Ari Pakman): a minus sign missing in the exponent
Page 315, equation (16.14) (Hao Bo): the boldface partial residual r
should be annotated differently, since over the page we use boldface
r for the full residual.
Page 320, in the regression boosting algorithm, in 2(a) (Ari Pakman): $F(x_i)$ should be $F^{b-1}(x_i)$
Page 322, in line 2 from the bottom (Ari Pakman): $O((p-r)rn)$ is missing the closing parenthesis.
Page 330, line 2 (TH): "three times" should be "1.58" times
Page 338, section "Shrinkage" (Chun Li): The blue ensemble uses a shrinkage parameter 25 times smaller, not "20 times."
Page 380, last line (Chun Li): "as opposed to", not "as apposed to"
Page 402, end of 2nd paragraph (Ari Pakman): the reference should be to figure 21.5, not 31.5
Page 406, third and fourth paragraphs (Ari Pakmna): the references to
the green dotted curve should be to the blue dotted curve; above
(20.30), $x^{4000}$ should be $x^{*4000}$
Page 415, equation (20.54) (Ari Pakman): $\beta$ should be
$\beta^{(b)}$. The same in the 2nd line below (20.55)
Page 419, in (20.66) (Ari Pakman): the first term should have $z^2$ instead of $z$.
A further 48 possible errors were reported by Yoshihisa IJIRI that he
and his team encountered in the Japanese translation of the book
Errata since corrected March 2017 online version and 4th printing, and corrected for December 2017 online version and 5th printing.
Page 9, line -8 (Ariel): "non-nsignificant" $\rightarrow$
"non-significant"
Page 10, Figure 1.5 (Chun Li): The t-statistics used here were actually the unequal-variance version (Welch); this will be replaced by the histogram using the pooled-variance
Page 27, Figure 3.2 (David Goldberg): the black curve (posterior wrt
uniform prior) is incorrect. What is plotted is the likelihood, which
integrates to 0.977. One needs to divide this function by 0.977, which
changes slightly the points of intersections
Page 40, equation (4.7) (Jui-Chung Yang): the $\sigma^2$ in the
denominator should be be $\sigma^v$
Page 49-50 (Chun Li): 3.13 is the unequal-variance (Welch)
t-statistic for gene 136; will be replaced by 3.01, the
pooled-variance t-statistic, to achieve consistency with Chapter
1. There were 26 exceedences, and hence a p-value of 0.0026
Page 67, before (5.60) (Douglas Rivers): "derivate" should be "derivative"
Page 68 bottom (David Goldberg): It says Lindsey's method will be
discussed in Chapter 15. It says the same thing in footnote 5 in page
171. But it is actually explained in Chapter 10 (page 171) and not in
Chapter 15. The index entry for Lindsey's method points to page 68, but it should be to page 171.
Page 78, 2 lines from the bottom (Jui-Chung Yang): "Table 6.2 going on
to show that three species were trapped 44 times each, and so on."
Should instead be "Table 6.2 going on to show that 44 species were trapped three times each, and so on"
Page 83, below (6.27)(Chun Li): better to be "as in (6.17)".
Below (6.28) : N is the total number of butterflies trapped (or the total number of words in Shakespeare's work), not the number of species
Page 105, above (7.48) (Harel Lustiger): $Z \sim \chi^2(\nu)$ should
be equivalent to $Gam(\nu/2,2)$ using the notation in table 5.1
Page 115, (8.19) (Douglas Rivers): $\alpha_j$ rather than $\alpha_i$
Pages 120 and 121, section 8.3: "486" should be "487"
Page 127, last line (Sheridan Grant): "the availability [of] massive ..."
Page 134 (in section 9.2)(Florian Krach): “The response for each patient is survival time in [days]” (instead of months).
Page 151 (Greenwoods formula)(Florian Krach): There should be $n_k \hat{h}_k \sim Bi(n_k,h_k)$ (instead of just $ \hat{h}_k \sim \ldots$ so that $var(\hat{h}_k) = h_k (1-h_k) / n_k)$. And in the next sentence: "Plugging in [$h_k$] ..." (instead of $\hat{h}_k$).
Page 167, line -3: t statistic is 3.01, not 3.13
Page 172, footnote 7 (David Goldberg): the factor $(1 - y_k)$ in the formula should be $(n - y_k)$
Page 184, Figure 11.2 (David Goldberg): The 0.953 should be $m(0.751)$. But $m(0.751) = 0.975$
Page 190, first sentence of section 11.3 (David Goldberg): reference
(11.22) should be (11.23)
Page 191, line -8 (David Golberg): endnote 4 and endnote 5 should
change positions and numbers in the endnotes
Page 202, line 5 (David Goldberg): (11.17) rather than (11.16)
Page 213, above equation 12.20 (Francisco Rodriguez Algarra): "as"
$\rightarrow$ "an"
Page 214, just after equation (12.22) (David Golberg): The reference
for err = 0.72 should be equation (12.11), not (12.9)
Page 215, figure 12.2 (Francisco Rodriguez Algarra): the different
error estimates are colored blue (apparent) or red
(cross-validated). In the main body of the text, however, these are
described as solid (apparent) or dashed (cross-validated) lines,
with no mention of the colors
Page 217, Figure 12.3 (David Goldberg): this figure is truncated at 2.0
on both axes (causing the pile-up). The next printing will replace
this with the non-truncated version.
Page 227, 3rd line of section 12.4 (David Goldberg): the reference
should be (12.19), not (12.16)
Page 284, table 15.1 (David Goldberg): deviances, when recomputed, are slightly off; should be 138.6, 137.0, 65.1,64.1,63.7
Page 288. figure 15.8 (David Goldberg): the points in the plot are
truncated at [-6,6]; the next printing will replace
this with the non-truncated version.
Page 341, Algorithm 17.4, part 2(a) (David Thaler): 1=1...n should be i=1...n
Page 345, first line of (17.15)(Sheridan Grant): missing a close parenthesis on the right-hand side
Page 360, 3 lines above Figure 18.5 (Paul on Discuss): "just under 0.093% errors"
should be "just under 0.93% errors"
Page 371 section 18.6 (Sheridan Grant): "Two early statistical references... are Ripley (1996) and Bishop (1995), [and] Hastie et al (2009) devote one chapter to the topic."
Page 376 section 19.1 (Sheridan Grant): "We see three different classifiers..., and they all classif[y] the points perfectly."
Also, we mention the +1 and -1 coding, but don't explicitly say y is coded as such.
Page 403, above (20.27) (Sheridan Grant): Too many *s here. Probably (20.27) wants no *s as well, nor does t(c,x*) below Figure 20.6
Page 465 (Michael Godfrey): "Teller" in author index points to page
261, but does not appear there. [authors' note: on page 261, the
"Metropolis" reference list is abbreviated, and (two) Tellers are in
this list. In the next printing, the index will also point to the
bibliography page where Teller appears (in this case 459), and for all
such similar occurrences.]
Errata since November 2016 corrections submitted for 2nd printing, and corrected in March 2017 online version.
Page 27, Figure 3.2 (Zepu Xi): Vertical axis is posterior, so label
should be $g(\theta | \hat \theta)$ using the notation in the chapter
Page 45, equation (4.34) (Ari Pakman): the mean should be $\theta$,
not 0. Also (twice) in the 4th line below (4.35)
Page 45, line -8 (Pablo Davalos): "How accurate is $r(\mathbf{x}, \mathbf{y})$"
should be: "How accurate is $r_{\mathbf{x}, \mathbf{y}}$", following the notation in line -10
Page 49, equation (4.42) (Josh O'Brein): the $t$ on the RHS of the
inequality should be $|t|$ (though in this specific application, with
$t=3.13$, it happens to makes no difference)
Page 55, below equation (5.4) (Ari Pakman): "$p$ small" should be "$\pi$ small
Page 69, (5.64) (Ari Pakman): the '$=$' in the lower right should be a '$-$'
Page 71, (5.74) (Qike Li): the first expression should be
$f_{\mu}({\bf S}) = \prod^{L}_{l=1}e^{-\mu_{l}}\mu_{l}^{S_{l}}/S_{l}! $
. (5.75): the $S!$ should be $S_+!$, and $x_l$ should be $S_l$.
Page 76, table 6.1 (Christoph Hanck): Last entry for "Formula
(6.7)"-row should be $7\cdot(1/4)=1.75$
Page 80, Eq (6.16) (Manuel Haussmann): should read "...($\theta t)^3/3! -$ ..." instead of "...($\theta t)^3/3! +$..."
Page 92, equation (7.12) (Yi Liu): in [...], the first $x$ should be $x_i$
Page 93, line -8 (Ari Pakman): "inadmissable" should be
"inadmissible"
Page 111, Eq (8.11) (Manuel Haussmann): the last product should read "$\prod_{i=1}^N$" instead of "$\prod_{1}^N$"
Page 115, last paragraph (Ari Pakman): the reference to eq. (7.19)
should be to (7.20)
Page 117, Table 8.4 (Ari Pakman): "binomial" should be capitalized, to
be consistent with Table 5.1
Page 117, eq. (8.25) (Ari Pakman): in the equation on the right, it
should be $\gamma(x_i'\alpha)$.
Page 118, second line below (8.28) (Ari Pakman): the two $\alpha$'s
should be $\hat{\alpha}$
Page 118, the line below (8.30)(Ari Pakman): in the variance of the
exact OLS result, the exponent of $\sigma$ should be -2, not 2. This
agrees with (8.30) (since $\Sigma$ is diagonal with entries
$\sigma^2$) and also with the fact that for linear regression $y \sim
N(x'\beta, \sigma^2)$, the coefficient $\alpha$ in (8.24) is
$\beta/\sigma^2$. This latter fact should probably be in a footnote, otherwise
this change might confuse.
Page 129, eq (8.53) (Ari Pakman): the subscript $\alpha$ is missing on
the left-hand side.
Page 130, below (8.59) (Ari Pakman): there is a $-\log$ missing before
the ratio of $f$'s
Page 141, eq (9.27) (Ari Pakman): the sums should be have "i=1"
instead of "1"
Page 146, end of second paragraph (Ari Pakman): reference should be to (9.32), not (9.31)
Page 146, 4 lines before Section 9.5: "estimate or" rather than
"estimate of"
Page 147, eq. (9.42) (Ari Pakman): the sums should be have "i=1"
instead of "1"
Page 154, equation (9.63) (David Holdberg): the sign in front of the integral should be +, not -.
Page 159, eq (10,16) (Ari Pakman): in the right, the sum should read
"b=1"
Page 205, equation (11.77) (David Goldberg): Remove the subscript 0 in
$\theta_0$ in this equation and two lines below
Page 218, line -4 (Ari Pakman): the first symbol } should not be
there
Page 224, eq (12.64) (Ari Pakman) the sum should be over b, not j
Page 254, Eq (13.73) (Andres): The denominator should have a plus
rather than a minus
Page 256, eq (13.79)(Ari Pakman): the sum should be over j, not i
Page 260, line 5 after (13.84) (Manuel Haussmann): doubled "the"
Pages 308-310 (Ari Pakman): the symbol $n$ (number of data points)
appears in Section 16.3 but disappears in 16.4, making the notation
inconsistent. For example, the equation in page 308, line -3 has a
factor $1/n$, while the same equation in page 309, line -4 (written as
$c_j(\lambda) = \lambda)$ lacks the factor $1/n$
Page 321, l4 of paragraph "Extensions of the Lasso" (Manuel
Haussmann): missing an apostrophe in "dont"
Page 403, second paragraph (Ari Pakman): the first $\bf x$ should be
$\bf x^*$.
Page 407, equation (20.32) (Noud van Giersbergen): there should not be
a $/n$ in the formula
Page 412, two lines below (20.42): replace the bold $\mu$ with
$\mu_i$
Page 420, line 5 below (20.68)(Ari Pakman): $Cov_x = T(x)$ should be $Cov_x = \nabla T(x)$
Page 422, line -5 (David Goldberg): The reference should be to (21.1) and (21.2) instead of (3.1) and (3.2)
Page 456, reference (Efron 2016) (Ari Pakman): reference should be
"Efron, B. 2016. Empirical Bayes deconvolution estimates, Biometrika, 103(1), 1-20"
Errata since August 2016 first printing. Second printing in
progress, corrected for all errata below.
Page 8, line 2 of Section 1.2: Should be 47 ALL and 25 AML
Page 11, Section 1.3 (Paul King): "...Tukey's explanatory/confirmatory
system" should be "... Tukey's exploratory/confirmatory system"
Page 16, above (2.12) (Andreas Buja): "distributuion" ->
"distribution"
Page 17, (2.19) (Patrick Li): Should be $\pm
1.99\cdot\widehat{\mbox{sd}}$
Page 32, line+7 (Ronald Wampler): change 0.083 to 0.074 (based on a
million simulations)
Page 37, line 13 (David Olive): should be $\Phi^{-1}(F_{100}(t_i))$.
Page 41, 3 lines above (4.12): we reference (3.6) which should be (4.6)
Page 49, line 3 (Brandon Greenwell): reference to Figure 1.1 should be
Figure 1.4
Page 52, (4.45) (David Olive): the second bracket of the three brackets needs a dx
Page 54, equation 5.2 (Romann Weber): Should read ${\cal
N}(\mu,\sigma^2)\sim \mu+\sigma{\cal N}(0,1)$. Applies to the
preceding text, and to the footnote below as well.
Page 59, (5.23): Not an error, but form similar to (5.21) more
suitable
Page 62, (5.37) (Romann Weber): instead of the parens, should be
horizontal division bar.
Page 70, lin -1: As should be bold
Page 71, line -3: $ce^{\alpha'_1 y} + \cdots$
Page 78, line -3 (Andreas Buja): "speciman"-> "specimen"
Page 79, line+2 (Andreas Buja): "speciman"-> "specimen"
Page 79, (6.12) (David Olive): $Poi(\theta_k t)$ (delete the
comma)
Page 102, (7.43) (Simon Minovitsky): should be $\hat{\beta}^T S \hat{\beta}$ instead of $\hat{\beta}^{-1} S \hat{\beta}$.
Page 107, (7.59) (David Olive): should be inverse on last parens
Page 165, (10.32): instead of the parens, should be
horizontal division bar.
Page 221, (12.50) (David Olive): $(y_i - \hat{\mu}_i)^2$ (need a hat
on the $\mu_i$)
Page 245, Table 13.3 (John Williams): 20-50 should be 20-150
Page 310, end of step 3(b) in algorithm 16.3 (David Olive): should be
$X_{\cal A}\delta$
Page 310, line -11 (David Olive): endnote 1 rather than 4
Page 411, line 17 (David Olive): "valus" should be "values"
Page 411, (20.41) (David Olive): the $\mu_1$ should be $\mu_i$
Page 431, 5th line of Fig. 21.3 (David Olive): "Tghe" should be
"The"