Archive for November, 2008

The Wheel keeps rolling…

Posted in Books with tags , , on November 29, 2008 by xi'an

For those of us who started reading The Wheel of Time series in 1990 when the first volume The Eye of the World appeared—I do remember reading it while house-sitting in Ithaca—, the vigil for the conclusion of this epic has been close to unbearable: eleven volumes and twenty-eight years later, the end is not yet in sight! When Robert Jordan died last year, it even sounded as if the investment in reading these (mostly enjoyable) seven thousand four hundred and nineteen pages (plus the three hundred thirty-six pages of New Spring!), remembering more than two thousands of characters and hundreds of places, had been in vain (except for the pleasure of reading the eleven first volumes!) as the last and final volume was too far from completion and that the one-before-last book, Knife of Dreams, was not pulling enough threads together… A few weeks after Robert Jordan’s demise, however, his wife and editor Harriet Rigney selected a young fantasy writer, Brandon Sanderson, to write the missing volume based on the numerous notes and indications Robert Jordan gathered and prepared before he died. From reading—and liking—both first volumes of the Mistborn series by this author, I gather this is potentially a good choice, a priori.

Unfortunately, Brandon Sanderson has just posted an announcement on his blog that he had now reached the 300k word landmark on his completion of the first draft of A Memory of Light. Unfortunately, because he added that it sounded more like 45%-50% of the book was now completed, rather than 75% as originally planned! Robert Jordan had already warned that this last volume would be the longest of all and that his publisher Tor would have to invent a new type of binding. But this most importantly means that the end of The Wheel of Time is not in sight, by far, and that we will most likely see the first half in print a year or so before the second half. In other words, the end is not due in one volume but in two, even if they share the same title… I do not suspect any foul play from the publisher trying to make more profit about this best-seller (who knows?, fans may dislike Mr. Sanderson’s rendering of Robert Jordan’s style and shun the second volume altogether). This additional delay is nonetheless frustrating, if only because one needs to re-read those earlier volumes from time to time to make sure to remember enough of the main characters and of the main threads in the Pattern.

The saga of publishing The Wheel of Time will thus span more than twenty years, a phenomenon quite in tune with the range and character gallery of the books. Two, maybe three generations, are together waiting for the final bit. The Wheel weaves as the Wheel wills, but most of us would like to see it grind a wee less finely towards the end and stop rolling at a foreseeable horizon…

The Sea Wall

Posted in Books with tags on November 29, 2008 by xi'an

The movie The Sea Wall by Rithy Panh, based on a novel by Marguerite Duras, now has a trailer posted on YouTube:

If I am posting an entry on this movie, it is not for being a particular fan of Marguerite Duras, but because, along with movie stars Isabelle Huppert, Gaspard Ulliel, and Astrid Bergès-Frisbey, it features my friend and coauthor Randal Douc whose many skills, besides Mathematics and Statistics, include play-acting and play-writing, with the impressive Les Hommes désertés, impressive by its intensity and by its raw exposure of the human soul.. The scope of Les Hommes désertés is far, far broader than trying to understand how such a thing as the Red Khmer genocide could have taken place. Even though clearly derived from this genocide, the text is universal and intemporal. The play was first created in Phnom Penh, before being played in Orléans two years ago. (The movie is expected to play in movie theaters in early January in France.)

The twilight zone!

Posted in Statistics with tags , , on November 26, 2008 by xi'an

I had the weirdest impression this morning when, while looking at entries on defensive sampling on Google, I ended up on a webpage that suspiciously looked like my own work! After getting over the first shock (it was way too early for shouting out loud!), I went up the directory hierarchy to understand at last that my entry into the collective book Computational Statistics by James E. Gentle, Wolfgang Härdle, and Yuichi Mori has a free version available on the Web, as all entries of this book have…. This available web-book can be quite useful as additional material when teaching an introductory course on computational statistics.

Bayesian reading groups in Paris

Posted in Statistics, University life with tags on November 24, 2008 by xi'an

There are two new Bayesian reading groups in Paris, one centred on model choices issues as explained here and another one focussing on WinBUGS. Attendance is free and open to anyone interested. I am not involved with the second group but try to attend the first one whenever I am free….

In the same spirit, Andrew Gelman will visit Engref in Paris on December 11-18 2008. He will give an advanced crash course on “Bayesian thinking” on December 12 and December 15, as well as a seminar on his book “Red State, Blue State, Rich State, Blue State: Why Americans vote the way they do?” at 2pm on December 15, still at Engref.

At last, there is a workshop on December 5 at Université Paris Ouest Nanterre-La Défense [Salle 614, Bâtiment G] on Random measures and Bayesian Nonparametrics organised by Patrice Bertail et Giovanni Peccati.

The approximation in ABC

Posted in Statistics with tags , , on November 21, 2008 by xi'an

An interesting new paper about ABC was posted by Richard Wilkinson on ArXiv today. The main point in the paper is to replace the approximation error resulting from the loose acceptance condition in regular ABC (i.e. when accepting a simulated value within a distance ε from the observation) with an exact ABC simulation from a controlled approximation to the target, essentially a convolution of the regular target with an arbitrary kernel π. The idea is indeed interesting in that the outcome is completely controlled, thanks to the degree of freedom brought by the choice of the kernel π, but I think its scope does not compare with the kernel smoothing perspective found in Beaumont et al. (2002). The convolution of the distribution f of the observables with the arbitrary kernel π indeed does not bring us closer to the ideal inference based on the true posterior, while the nonparametric motivations of Beaumont et al. (2002) and more recently of Blum and François (2008) have the goal of improving the approximation to the true posterior, based on previous simulations. (The criticism in the current paper of the Epanechnikov kernel as a poor choice for a measurement error distribution is thus off-key, given that the incentive is nonparametric in Beaumont et al. (2002). Note also that the final extension of the paper both to Monte Carlo integration and to model choice is quite indistinguishable from a nonparametric approximation of the posterior distribution.) In that sense, both papers of Beaumont et al. (2002) and of Blum and François (2008) develop adaptive methods in the sense of sequential Monte Carlo, like the PMC version of ours.

This type of work also relates to the theory of model approximation built by Tony O’Hagan in the early 2000’s (as pointed out in the paper).

My main criticism of the paper is that the approach proposed by Wilkinson requires a modification of the ABC algorithm and thus that the algorithm is exact only after this modification, i.e. after changing the problem. The algorithm also includes an upper bound c on the convolution kernel π, which is not an enormous requirement given that π is arbitrary but still is a requirement. The extension of the method to include MCMC steps as in ABC-MCMC is altogether not surprising, but the comparison of two MCMC versions is rather interesting, one avoiding the use of the upper bound c by integrating both approximations in a single acceptance step. The paper rightly concludes by stating that more work is needed, in particular to include the additional approximation step due to the use of summary (insufficient) statistics, which is much more of an issue, as discussed in yesterday’s blog.

Sufficient statistics for ABC

Posted in Statistics with tags , on November 20, 2008 by xi'an

The issue of picking quasi-sufficient statistics in ABC algorithms is quite important and Joyce and Marjoram have published (last August, sorry I’m late!) a proposal in Statistical Applications in Genetics and Molecular Biology that is addressing this issue, at least to the extend that the inclusion of a new statistics within the set of summary statistics can be assessed in terms of likelihood ratio. There are difficulties with the method, first and foremost that the construction of the statistics is not discussed, while the method is not independent from parameterisation, and also that the order in which the statistics are considered is determinant for their inclusion/exclusion. A regularisation of the method proposed at the end of the paper is to use a forward-backward selection mechanism. This does not address another issue, namely the impact of the correlation between the summary statistics. Note at last that the method still depends on an approximation factor that needs to be determined by the statistician/experimenter running the algorithm.

Chronicle of a disaster foretold

Posted in Running with tags , on November 16, 2008 by xi'an

Today was the “big day”, i.e. the day of the half-marathon in Boulogne-Billancourt, which is my favorite race of the year because it is flat (the only gradients are when passing under other roads), fast, mostly protected from headwinds (even though it borders the Seine river), well-attended (more than 4500 runners this year) without being too crowded like the Paris half-marathon and well-supported with several bands along the race, including a Breton piper band! The winner this year, Nicolas Manza, won in the incredible time of 1:00:09. He was 1 minute and 20 seconds ahead of the second runner… Each time we passed the head group, he was clearly far far ahead of the other runners. This is still more than one minute above the World record, but it shows how fast the race can be.

My own time is obviously much less glorious since I ended up in 1:21:06, almost two minutes more than my record time of 1:19:13 last year and my worst time on that race. As suggested by the title of this post, I knew beforehand this was not going to be a top race. Since my half-marathon in Argentan, I have indeed recovered very slowly, with recurrent pain in the upper right tight, repeated bouts of mild tendinitis under the right knee and even two rounds of stiff necks that prevented me from running for several days in a row… In total, I had to skip more than two days a week of running and, worse, all my training days! Since Argentan, the only training day I have had is a series of 300 meters. So, despite running almost every day over one hour for the past 10 days, I knew I had not enough training to keep with the pace of last year. The beginning went on fine, actually, with a rythm of about 3:46-3:50 per kilometer, and I kept the pace till the 16th km. But then the 1:20 flag runner passed me along with a dozen runners and I realised I was in trouble when I could not keep up with them! I then ended at a pace closer to 4:00 per kilometer with a group of runners that I joined almost from the start, even though I had some reserves left to run the last kilometer a wee faster and to overcome two runners. I now need to fully recover….