Bayesian Core

A Practical Approach to Computational
Bayesian Statistics

J.-M. Marin & Ch. P. Robert

Springer-Verlag [site], New York, 2007, ISBN 0-387-38979-2

Typos in the second printing, to be corrected in the third printing

Footnote 7 in Chapter 1 points out the wrong website for CRAN, the correct URL is http://cran.r-project.org/ (Thanks to Lawrence Joseph).
In the analysis of normaldata, page 19, x ("x bar") should be defined as the empirical mean over the 90 observations and σ² is taken as the empirical variance.
The power of (σ²) on line 5, page 23 (before formula (2.7)), should be -λ_σ -3/2, not -λ_σ
The sufficient statistic s²_x in (2.7) page 23 is defined as the sum of squares
On Figure 2.6, τ (not t) should label the horizontal axis (Thanks to Jarrett Barber and U. Wyoming STAT 5680).
On page 35, S should be S² in two places below the expression for B₁₀^π (Thanks to Jarrett Barber and U. Wyoming STAT 5680).
On page 38, in Exercise 2.22, x̄+ȳ should be x̄-ȳ (Thanks to Jarrett Barber and U. Wyoming STAT 5680).
On page 52, in the comment, σ² should multiply ω_i,i in two places (Thanks to Jarrett Barber and U. Wyoming STAT 5680).
On pages 52 and 62, the indeces n-k-1 and n of the t distribution denotes the degrees of freedom, while in other places in the book, it denotes the dimension. The correct notation should be T(n-k-1,0,1) and T(n,0,1) (Thanks to Jarrett Barber and U. Wyoming STAT 5680).
On page 56, in second sentence before Exercise 3.5, M should not be inverted (Thanks to Jarrett Barber and U. Wyoming STAT 5680).
On page 57, in Exercise 3.8, (q,p) should be (q,k+1) and in the final paragraph the reference to Table 3.1 concerns β₀, not β₁ (Thanks to Jarrett Barber and U. Wyoming STAT 5680).
On page 59, the power of (σ²) in the middle of the page is -(k+1)/2, not -k/2 (Thanks to Xiaohui Chen).
On page 60, the posterior variance of β involves a (n-2) in the denominator, not an n, and this implies a multiplication by 33/31 of the last column of both Table 3.3 and Table 3.4. In fact, the scale matrix of a T distribution is not the variance! The corresponding R programs have been corrected to this effect (Thanks to Xiaohui Chen).
On page 64, the first c+1 in the expression of the marginal for H₀ should be c₀+1 ((Thanks to Jarrett Barber and U. Wyoming STAT 5680).
On page 64, in the window section, the test is on β₇ = β₈ = 0, not β₈ = β₉ = 0 since the results and code in progs.3.txt correspond to omitting covariates x₇ and x₈, and the model has an intercept (Thanks to Jarrett Barber and U. Wyoming STAT 5680).
On page 66, the quantile to be considered in the remark on the confidence interval is for (n-k-1) degrees of freedom, not for (n-k). See Exercice 3.15 for an explanation (Thanks to Xiaohui Chen).
On page 68, there is a mistake similar to the one pointed out for page 60: c/(n(c+1)) in the posterior variance should be c/((n-2)(c+1)). This has again consequences on Table 3.6. There is also an extra lower case x after X at the end of the first equality in the expression of the posterior variance (Thanks to Jarrett Barber and U. Wyoming STAT 5680).
On page 74, the 0.05 in the standard deviation argument to the R code rnorm function should be 0.1 (Thanks to Jarrett Barber and U. Wyoming STAT 5680).
The unit vector 1_n on page 78 should be 1_k (Thanks to Brunero Liseo).
The projection β_γ at the top of page 79 is incorrectly defined. There is an X missing in front of β, even though the projection is correctly defined in the R programs (Thanks to Xiaohui Chen).
On page 79, the prior π(γ|X) should read π(γ) to avoid confusion with posterior (Thanks to Brunero Liseo).
In the remark page 89, the log-odds ratio should be log{p_i/(1-p_i)} (Thanks to Lawrence Joseph).
On page 90, there is an equality instead of a proportionality term in expression (4.2) of Example 4.2 (Thanks to Jarrett Barber and U. Wyoming STAT 5680).
In Exercise 4.7, page 93, scale=6.2 should be rate=6.2 (Thanks to Amandine Pierrot).
On page 100, in Exercise 4.12, y_i should be defined as the indicator that z_i positive_ as in Example 4.2 (Thanks to Jarrett Barber and U. Wyoming STAT 5680).
On page 110, the Poisson likelihood uses 1/μ_i! instead of 1/y_i! (Thanks to Amandine Pierrot).
On page 153, in Exercise 6.3, σ_i should be σ_j in the conditional normal prior and the empirical variance is missing a square power (Thanks to Jarrett Barber and U. Wyoming STAT 5680).
At the bottom of page 154, the conditional distribution of z also conditions on p (Thanks to Xiaohui Chen).
On page 165, in Algorithm 6.4, there is a parentheses problem in the expression for τ in that τ only deal with the first pair (Thanks to Jarrett Barber and U. Wyoming STAT 5680).
On page 178, there is a superfluous ")" in the index of p_j(k+1) and a missing ² in σ_(j+1)(k+1) in the definition of the split move (Thanks to Xiaohui Chen)