********************************************************************** Errata for Bayesian Computation With R ********************************************************************** Updated July 25, 2008 Most of these errors have been corrected in the 2nd printing, June 2008. The errors marked with * were found after the 2nd printing. ********************************************************************** 1. page 1, line 6. Should be in an introductory statistics class” *2. page 47, line 8. The lines command should be > lines(theta,1/tau*dnorm((theta-mu)/tau),lwd=3) 3. page 54, exercise 2, part (a). Should be "if we transform ..., then theta has a gamma density" 4. page 58, first full paragraph. The R output should read as follows: > data(marathontimes) > attach(marathontimes) > mycontour(normchi2post, c(220, 330, 500, 9000), time) > title(xlab="mean",ylab="variance") (the mycontour function was assigned to a variable d actually, the mycontour function doesn't return any output) 5. page 73, exercise 5, line 7. Should be "Suppose that (a, b) are assigned the noninformative prior proportional to 1/(ab)^2." Also the posterior expression for (theta1, theta2) in the middle of page 73 should have an extra factor of 1/(ab). 6. page 73, exercise 6, line 6. Should be "a series of newspaper ads is run in the two areas for a period of four weeks" 7. page 74, second line. Change .417 to 2.4 and (100, .4) to (100, 2.5). 8. page 77, line 13 from bottom. Replace "the prior density. The logarithm" with "the prior density g(theta). The logarithm" 9. page 83, the R output at the bottom of the page should be replaced by > se=sqrt(diag(fit\$var)) > fit\$mode-1.645*se [,1] [,2] [1,] -7.281247 5.663958 > fit\$mode+1.645*se [,1] [,2] [1,] -6.35634 9.485062 The last paragraph should read ... So a 90% interval estimate for logit(eta) is \$(-7.28, -6.36)\$, and a 90% interval estimate for log K is (5.66, 9.49). 10. page 89, last line. The standard error of the importance sampling estimate should be sqrt((sum(h(thetaj) - hIS) w(thetaj))^2)/sum(w(thetaj)) (the square root should extend only on the numerator term) 11. page 91, line 10 from bottom. Should be "sampling importance resampling" 12. page 94, 7th line from bottom. Should say ... "use of rejection sampling" 13. page 102, third paragraph, line 8. Change wP = w to wT = w. 14. page 110. There is a small mistake in the coding of the R function groupeddatapost.R. One can correct the listing on page 110 by changing the line of code z=z+freq[nbins]*log(1-pnorm(cpoints[nbins],m,s)) to z=z+freq[nbins+1]*log(1-pnorm(cpoints[nbins],m,s)) With the use of the correctly coded function, the results in Section 6.7 and 6.8 will change slightly (the revised printing will have different output on pages 111 through 115). 15. page 128, R code at bottom. Insert before the first line of code > p=exp(s\$par[,3]) 16. page 142, 13th line from bottom. It should say "of transplant deaths" 17. page 143, line 2. In the R input line, lam should be lambda. 18. page 155, line 13 from bottom. > ys24=rpois(1000,e[94]*lam94) should be replaced by > ys94=rpois(1000,e[94]*lam94) 19. page 158, ninth line from bottom. Should be "normal density with mean mu and standard deviation sigma" 20. page 174, line 3 before the last paragraph. The last parenthesis in the R code should be deleted. 21. page 180, line 11 from bottom. To get the R output as shown, the command > cbind(logK,logBF,exp(logBF)) should be replaced by > cbind(logK,logBF,exp(logBF))[seq(1,26,by=5),] 22. page 188. In the last expression before Section 9.2.3, the transpose operator should be ' instead of T. 23. page 208, line 6 from bottom. PEAK should be defined as PEAK = beta0 - beta1^2/(4*beta2) also in line 4 of exercise 3, , "if beta3 < 0" should be replaced with "if beta2 < 0". 24. page 209, first line. PEAK AGE should be defined as PEAK AGE = -beta1/(2 beta2) 25. page 217, line 26. The use of solve is not efficient in this case. The calculation is better done as: mn=solve(t(x)%*%x,t(x)%*%z) 26. page 219, second paragraph, line 2. The second percentile 25th should be changed to 50th. 27. page 220, line 3 from bottom. Line "41-50", + "31-40", "21-30") should be replaced with "41-50", "31-40", "21-30") 28. page 235, formula at bottom of page. yi in the first line of the expression should be replaced by xi 29. page 236, exercise 4, line 5. In the inequality, p_n should be replaced by p_N. 30. page 247. In the R code at the end of Section 11.5, the R command > plot(sqrt(perot),sqrt(buchanan)) should precede > lines(xo,meanp[2,]) 31. page 251, fourth line. The line containing the single character b should be replaced with beta[i , 1:3] ~ dmnorm(mu.beta[], R[ , ]) 32. page 255, 6th line. Should be "Fitting a binomial/beta" ACKNOWLEGEMENT: Thanks to David Ardia, Josue Guzman, Peter Lee, and the students of MATH 648 for finding these errors.