Archive for JSM 2009

Newer Entries »

JSM 2009 impressions [day 2]

Posted in Books, Statistics, University life with tags , , , , , , , , , , on August 4, 2009 by xi'an

Julien Cornebise wrote his impressions on yesterday [day 2] as comments to day 1 and he is welcome as a guest editor! I completely agree with his views on George Casella’s Medallion Lecture on design, which emphasized the need to reconsider this somehow neglected part of the Statistics curriculum. George’s lecture was both passionate and broad, which made it accessible to the large audience there. It was based on his Statistical Design book, on sale at the Springer Verlag booth in the Exhibit hall when you go to check the Enigma machine at the NSA booth. Along with a whole table of new books in the Use R! series, soon to be augmented by our book Introducing Monte Carlo Methods with R with George Casella, which is available in a draft version at the booth. (We actually signed the contract for Introducing Monte Carlo Methods with R with Springer yesterday afternoon.) The Springer editor, John Kimmel, is one of the ASA Fellows this year, in recognition of his support of the dissemination of new ideas in Statistics (my wording) and this is a great initiative from the ASA committee on Fellows as he unreservedly deserves it, if only for launching the Use R! series!

PICT6678

As mentioned by Julien, the session on the future of Statistics was reserved to the happy “fews” who managed to get a seat and others had to stay in the “present” thanks to this safety regulation that seems to be implemented on some talks/rooms and not others. I passed the first people being stopped by a fierce guard on my way to the “past”, ie to the cosmology and astrophysics session. There, I enjoyed very much Larry Wasserman’s talk on Nonparametric estimation of filaments for uncovering a challenging problem as well as for his elegant resolution of the problem. As well as the presentation by Laura Cayon of Detection of weak lensing, where I discovered that my old Purdue friend Anirban das Gupta was also involved in cosmology. I also went to the Monte Carlo and Sequential Analyses: Methods and Applications session, organised by Mike West, but the talks were too short to make much of an impact on me, even though I appreciated the talk by Minghui Shi on Particle stochastic search for high-dimensional variable selection that linked with Nicolas Chopin’s early work on exploring a large dataset and I was also intrigued by the talk of Ioanna Manolopoulou on Targeted sequential resampling from large Data sets in mixture modeling for using proxies to the real mixture model. The day ended up with a Board meeting for ISBA, that unfortunately took place outside in a hot humid weather… I now have to get ready for the Gertrude Cox Scholarship 5k race, since it starts at 6:15am (yes, am!).

Leave a comment »

JSM 2009 impressions [day 1]

Posted in Statistics, University life with tags , , , on August 3, 2009 by xi'an

Last afternoon, I attended the Bayesian model choice invited session, with talks by David Madigan, Hani Doss and Hugh Chipman. The room was packed and the talks were quite interesting. Madigan et al’s Sequential Bayesian Model Selection uses Mike West’s dynamic parameters to allow for more adaptive models, with the drawback of a lack of stabilisation in the estimation of the parameter which reflects the time-varying modelling. (I could not really understand why the model choice approach could not handle the hidden Markov chain of the model labels, though.)

The second talk, Doss’ Estimation of Large Families of Bayes Factors from Markov Chain Output, was more central to my interests (and to my own talk). and had several innovations worth mentioning. One was the use of the Bayes factor for prior comparison, which is something I never thought of before. In a strictly Bayesian perspective, this is not surprising as it means using the data to compare your priors! Completely un-orthodox! At a deeper level, I am still wondering at the validation of the Bayes factor in this setting… The second innovation was that, thanks to the same likelihood appearing in both numerator and denominator, Hani Doss was able to use a new type of bridge sampling estimator due to the identity

\mathbb{E}_{\pi_2(\cdot|x)} \left[ \dfrac{\pi_1(\theta)}{\pi_2(\theta)} \mid x \right] = \dfrac{m_1(x)}{m_2(x)}

and thus, using a posterior sample from one posterior is sufficient for approximating the Bayes factor in this case. A third innovation was using the control variates

\dfrac{\pi_i(\theta) - \pi_j(\theta)}{\pi_1(\theta) f(x|\theta)}

since they always are unbiased estimators of zero. A point I need to investigate further is how using improper priors (the example in Hani Doss’ talk was based on Zellner’s g-prior) is possible in this regard.

The talk by Hugh Chipman was about the BART (Bayesian additive regression tree) “machine learning” technique of Chpiman, George and McCulloch, which is always impressive as a completely non-parametric regression method. The innovation was in using many trees (50,200) simultaneously in the model, not as in model averaging, but as a basis for more complex dependences. The differences in the influence of the covariates seems to vane away as the number of trees goes up, but, as discussed during this session, this may be due to the frequency of uses of a covariate being too crude an indicator.

I then went to Aad van der Vart’s Le Cam lecture, on Some Frequentist Results on Posterior Distributions on Infinite-Dimensional Parameter Spaces, who, as usual, managed to give a very methodical and clear overview of the results on the asymptotics of Bayesian non-parametric estimators.

The meeting is, as forecasted, a monster of a meeting, but the conference center is so huge that it is not overwhelming. The JSM staff is quite efficient and they managed to solve the issue of the Series B editors meeting almost immediately.

http://www.amstat.org/meetings/jsm/2009/onlineprogram/index.cfm?fuseaction=activity_details&activityid=&sessionid=205089

5 Comments »

JSM 2009 [talk]

Posted in Statistics with tags , , , , , on August 3, 2009 by xi'an

I made cuts in my talk during the flight to Washington and here is the edited version, with only novelty the example of harmonic mean estimator on which I posted earlier this week.

Of course, there are way too many slides… This means I will presumably cut down on the ABC part while making the connection with Simon Tavaré’s talk through ABC-PMC.

1 Comment »

JSM 2009

Posted in Statistics, University life with tags on August 1, 2009 by xi'an

When looking at the programme of the Joint Statistical Meeting next week, I realised I had given “Computational Methods for Bayesian Model Choice” as the title of my talk in the Advanced Monte Carlo session:

Wednesday August 5, 2009 - Session 495 - Room CC-143A
Advanced Monte Carlo Methods—Invited
Section on Statistical Computing, Section on Bayesian Statistical Science, Interface Foundation of North America
Organizer(s): Yuguo Chen, University of Illinois at Urbana-Champaign
Chair(s): Yuguo Chen, University of Illinois at Urbana-Champaign
2:05 p.m. Computational Methods for Bayesian Model Choice—Christian P. Robert, Université Paris Dauphine; Jean-Michel Marin, University Montpellier II
2:35 p.m. Auxiliary Variable MCMC with Applications in Protein Structure Modeling—Jun S. Liu, Harvard University; Kou X. Sam, Harvard University
3:05 p.m. Approximate Bayesian Computation: What, Why, and When?—Simon Tavaré, University of Southern California
3:35 p.m. Floor Discussion

So I guess I will condense the talk I gave at MaxEnt 2009, including the latest remarks about the harmonic mean estimator and also presenting our ABC model choice approach to make a link with Simon Tavaré’s talk. I will also chair a SAMSI session on population Monte Carlo and SMC organised by Julien Cornebise, where Jean-Michel Marin will talk of our AMIS paper and Darren Wraith of the PMC for cosmologists paper.

Leave a comment »

Newer Entries »