Errata corrige

Bayesian Methods in Health Economics

Published

April 1, 2019

Warning

Errata corrige

Despite our best effort, there are a few typos or imprecisions in the published version of Bayesian Methods in Health Economics. We report those we have identified and welcome (well… we hope there won’t be many, but you know…) notification of any other mistakes (that can be highlighted by emailing me).

Page 21. The interpretation of the cost-benefit analysis is incorrect. In fact, the favourite option should be \(a\) (and not \(b\), as reported in the book), since the cost for Option \(a\) relative to the cost of hospitalization without vaccination is smaller, which is preferable. This is due to successive revisions to this examples (in the previous version the result was the other way around and the text was not updated properly).
Page 38. There is a typo in the equation for Bayes’ rule to compute the posterior probability \(\Pr(D\mid +)\). The correct calculation should be \[ \Pr(D\mid +) = \frac{0.99\times 0.1}{0.99\times 0.1 + 0.05\times 0.9} = \frac{0.09}{\boldsymbol{0.1485}} = 0.667 \] (so the correct value is 0.1485 — not 1.135 as incorrectly reported in the book).
Page 45, figure 2.5: the bottom part of the decision tree shows the possible outcomes when the operation does not go well. However, the headings for the branch out of the “Patient lives” random node are both “No”. The top one should read “Yes” (much as in the top half of the tree).
Page 105. The formula for \(\Delta\mbox{DIC}_h\) should be \[\Delta\mbox{DIC}_h = \mid \min_h (\mbox{DIC}_h) - \mbox{DIC}_h \mid\] (this makes a difference when the DIC is negative and without the absolute value operator, the ordering associated with the weight is wrong).
Page 120. There is a slight inconsistency between the code used to run the model

model = jags(data=data,inits=inits,parameters.to.save=params,
             model.file=filein,n.chains=2,n.iter=10000,n.burnin=4500,
             n.thin=1,DIC=TRUE)

and the text presented later on page 121, where it is stated that “… we consider 9500 simulations with a burn-in of 4500 for 2 chains”. The correct command replaces n.iter=10000 with n.iter=9500.

Pages 136-137. Again, there is a slight inconsistency in the code presented on page 136, where I have used n.burnin=4500 and n.thin=91 and the text and output presented on page 137, in which n.burnin=9500 and n.thin=81.
Page 147. In the call to bcea (approximately one third one the page), there is a reference to a (now) deprecated argument Ktable. I found this was largely irrelevant and thus removed it from the code. Everything else in this bit of code does work correctly, though.
Because of the changes in the R2jags package, the code in chapter 4 and 5 should be slightly modified to say attach.jags() instead of attach.bugs() (see more details here)