(by Julien Cornebise)
6am London tube, 7am Heathrow airport, 1pm (+5 timezones) JFK airport, 6pm Miami airport: Joint Statistical Meeting (JSM) 2011, here we come! I will have the pleasure to be one of your guest bloggers, very kindly invited by the editor Andrew Gelman for the duration of the meeting — you will find all our posts on The Statistics Forum grouped under the Category JSM2011.
Although technically JSM has not yet started, let’s put those lovely airport layovers to good use by warming-up.
With roughly 5,000 attendants and up to 48 parallel sessions, JSM 2009 (Washington DC) and JSM 2010 (Vancouver, Canada) were a treasure trove for statisticians, from the most theoretical French scholar to the most data-centric industry practitioner, and everyone in between. The price to pay: an excruciating one-week-long dilemma. Which talks to attend? — or perhaps more useful, how to choose? In such a large setting, even session-hopping won’t save the day — when it is possible at all.
The bulk of JSM: 20-minutes talks sessions. What to do with those? 20 minutes is most often just enough for an extended abstract with catchy graphs, which although a great way to advertise, cannot bring you much detailed information if you are already “in the know”. Plus, with 48 sessions at one single time, choosing will be simply impossible.
So one option that I tried the former years, and that I’m likely to try again this year: while we’re in Statistics heaven, let’s broaden our minds and pick sessions unrelated to our expertise themes! Where else, what other time in the year, will we have the occasion to taste by ourselves a completely different branch? Where else can the aforementioned die-hard Central-Limit-Theorem-on-Banach-Spaces-ist get a look at probabilistic computer CPUs? Where else can the government statistician taste why some think measure-theoretic stochastic approximation are important? And all of this is free of boredom! If, of course, we are all (most, hopefully) is in our own preferred field, no fear of dying in case of disappointment: one hour and it’s over. The only risk is to choke off in indignation if we forget to come with an open mind. For example in JSM 2010 I had the chance to attend the summary of a report to the Congress about “Data Mining in the War on Terror” — or why as an alien I have to give so many details to TSA (<unsurprising spoiler>bottom line of the report: , it’s a load of bollocks</unsurprising spoiler>): I would not do this every day, I would miss maths, but it was a fascinating insight, way more accurate that any newspaper report.
Of course there are some straightforward choices: Saturday and Sunday are mostly registered courses and workshops, booked in advance — e.g. I’m looking forward to hear any feedback about the Writing Workshop organized by NISS — having had the delight to spend one year in its co-hosted institution SAMSI, I have learned quite some tricks there, and am sure many more are to be gained at this workshop. I also honed there my presentation tricks, and hence am much interested in the Presentation Skills Workshop offered by ASA to JSM speakers.
Other even more straightforward choices are the staples: invited sessions by prestigious speakers that few would want to miss: presidential invited addresses, COPPS Fischer, Deming, Neyman lectures, Wald series, IMS Medallion Lectures… I am especially looking forward to those latter — I’m afraid my Bayesian bias might slightly show, there. Think of it:
- Michael I. Jordan on Applied Bayesian Nonparametrics (I have a thing for distributions on the space of distributions!);
- Qi-Man Shao on Stein’s Method — although a very theoretical topic, sometimes arid, I do hope to finally understand it one day; leftover trauma from an enjoyable reading group in Duke that I failed to wrap my head around;
- Jianqing Fan on “ultrahigh dimension”, to stay up to date about what is “ultrahigh” these days: when I was young (brhm hm… say, undergrad), it used to be a few hundreds; I’ve got a feeling the goalpost has slightly shifted … (kudos to Machine Learners around here);
- Sylvia Richardson on Bayesian methods and regression structures discovery applied to Health Sciences: a perfect resonance to the memories of last year’s program on Bayesian methods in Pharmacokinetics/Pharmacodynamics, which featured very rich debates about incorporating covariates;
- and finally, to stay in the same vein, Chris Holmes on Bayesian Analysis in High-Throughput Genomics: a key tool for Pharmacogenetics.
But again, for the rest of the week, JSM is by no means a small workshop between experts where papers get done around a drink. It is the best, the unique occasion I know of to see so broadly and easily what’s around us. We have all the evenings in the week, the mixers (e.g., shameless ad, the SBSS mixer or receptions by alma maters), to meet our direct field-colleagues and exchange while watching lovely Miami Beach sunsets. but for the rest of the day,
So for one week let’s get off the yellow brick road, and roam around, discover, broaden, “serendipit”, risk getting bored and win getting thrilled like nowhere else: that’s why I came, and although you’ll of course see me in some of the theoretical and methodological sessions that I wouldn’t miss, we will probably meet in the most unexpected ones too. Next connection: JSM Online Program — (while waiting for likely “Secondary check” at the border — Congress did not pick much out of that report). And if you have any sessions to suggest or other ways to plan your week, please leave a comment below while you’re converging to Miami Beach!