Typos in Chapters 6-7

Over the weekend and during the R exams, I managed to complete the solution set for Chapters 6 and 7 of “Introducing Monte Carlo Methods with R”. Chapter 6 only exhibited a few typos, despite me covering most exercises in Chapter 6, hence the merging of both chapters.

— in Exercise 6.13, both $\alpha$ and $\beta$ must use a double exponential proposal in the Metropolis-Hastings algorithm of question b,

in Exercise 6.15, the $\mathcal{L}(0,\omega)$ distribution should be a $\mathcal{N}(0,\omega)$ normal distribution,

— in Example 7.3, part of the code is wrong: it should be

> sigma2=theta=rep(0,Nsim)                  #init arrays
> sigma2[1]=1/rgamma(1,shape=a,rate=b)      #init chains
> B=sigma2[1]/(sigma2[1]+n*tau2)
> theta[1]=rnorm(1,m=B*theta0+(1-B)*xbar,sd=sqrt(tau2*B))

instead of

> sigma=theta=rep(0,Nsim)                  #init arrays
> sigma{1}=1/rgamma(1,shape=a,rate=b)      #init chains
> B=sigma2{1}/(sigma2{1}+n*tau2)
> theta{1}=rnorm(1,m=B*theta0+(1-B)*xbar,sd=sqrt(tau2*B))

(I frankly don’t understand where those curly brackets came from!)

— in Example 7.6, I forgot to include the truncation probability $\Phi(a-\theta)^{n-m}$ in the likelihood (!) and the notations are not completely coherent with Example 5.13 and 5.14 in that the x‘s became y‘s…

— in Exercise 7.21, rtnorm is missing sigma as one of its arguments.

— Exercise 7.23 has nothing wrong per se but it is rather a formal (mathematical) exercise

— in Exercise 7.25 the $x+a$ in question a should be $(x+a)$ to avoid any confusion.

This entry was posted on January 15, 2010 at 12:08 am and is filed under Books, R, Statistics, University life with tags Introducing Monte Carlo Methods with R, MCMC, Monte Carlo methods, simulation, typos. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.

5 Responses to “Typos in Chapters 6-7”

Alessandra Says:
January 18, 2010 at 11:28 am

:) !!!
Brute-force solution, O(1): try it on those curly brackets!

Reply
Alessandra Says:
January 17, 2010 at 5:31 pm

Overall, I think the use of “n” to indicate parameters or variables should be avoided in R codes. In fact, if ever one wants to use the debugger (not with a one line code of course!), “n” would be straightforwardly interpreted as the command “next” (instruction), and there would be no way to retrieve information about the “n” variable/parameter used in the program.

I know that most talented people never need using debug(), but I speak for us, humans…. :)

Reply
- xi'an Says:
  January 18, 2010 at 9:08 am
  
  I use spray for bugs…
  
  Reply
Colin Gillespie Says:
January 15, 2010 at 11:28 am

In the first code sample above, I think there’s a typo:

theta[1}]=rnorm(1,m=B*theta0+(1-B)*xbar,sd=sqrt(tau2*B))

should be

theta[1] = …

Also

B=sigma2[1]/(sigma2{1}+n*tau2)

should be

B=sigma2[1]/(sigma2[1]+n*tau2)

It’s those curly brackets again;)

Reply
- xi'an Says:
  January 15, 2010 at 11:44 am
  
  Typos within the typos! This is endless… Thank you very much, I will keep you posted about those curly brackets!!
  
  Reply

Xi'an's Og