Robin Ryder’s interview

Robin Ryder—with whom I am sharing an office at CREST, and who is currently doing a postdoc on ABC methods—, got interviewed in the March issue of La Recherche. (The interviewer was Philippe Pajot who wrote “Parcours de mathématiciens”, reviewed in a recent post.) The interview is reproduced on Robin’s blog (in French) and gives in a few words the principles of Bayesian linguistics. This two-page interview also includes a few lines of a technical entry to MCMC (called Monte Carlo Markov chains rather than Markov chain Monte Carlo) that focus on the exploration of huge state-spaces associated with trees. Overall, a very good advertising for MCMC methods for the general public through the highly attractive story of the history of languages…

Incoherent phyleogeographic inference [reply]

“Logical overlap is the norm for the complex models analyzed with ABC, so many ABC posterior model probabilities published to date are wrong.” Alan R. Templeton, PNAS, doi:10.1073/pnas.1009012107

Our letter in PNAS about Templeton’s surprising diatribe on Bayesian inference is now appeared in the early edition, along with Templeton’s reply. This reply is unfortunately missing any novelty element compared with the original paper. First, he maintains that the critcism is about ABC (which is, in case you do not know, a computational technique and not a specific statistical methodology!). Second, he insists on the inappropriate Venn diagram analogy by reproducing the basic identity

(presumably in case we had lost sight of it!) to argue that using instead

is incoherent (hence rejecting Bayes factors, Bayesian model averaging and so on). I am not particularly surprised by this immutable stance, but it means that there is little point in debate when starting from such positions… Our main goal in publishing this letter was actually to stress that the earlier tribune had no statistical ground and I think we achieved this goal.

Welcome, Robin!

Robin Ryder started his new blog with his different solutions to Le Monde puzzle of last Saturday (about the algebraic sum of products…), solutions that are much more elegant than my pedestrian rendering. I particularly like the one based on the Jacobian of a matrix! (Robin is doing a postdoc in Dauphine and CREST—under my supervision—on ABC and other computational issues, after completing a PhD in Oxford on philogenic trees for language history with Geoff Nicholls. His talk at the Big’MC seminar last month is reproduced there.)

And, in a totally unrelated way, here is the Sudoku (in Le Monde) that started my post on simulated annealing, nicely represented on Revolutions. (Although I cannot see why the central columns are set in grey…) I must mention that I am quite surprised at the number of visits my post received, given that using simulated annealing for solving Sudokus has been around for a while. Even my R code, while original, does not compete with simulated annealing solutions that take a few seconds… I thus completely share Dirk Eddelbuettel‘s surprise in this respect (but point to him that Robin’s blog entry has nothing to do with Sudokus, but with another Le Monde puzzle!)