(by Julien Cornebise)

Things have now officially started in the shiny city of Miami Beach. Statisticians shuffling from corridor to corridor, running from session to session, the $5 (! talk about debt ceiling…) convention-center-coffee in their hand, nametags around the neck (bravo to JSM team for making those double-sided!), going to Ballroom A instead of Junior Ballroom A (at the completely opposite end of the center), etc: quite a contrast with yesterday’s pre-meeting ghost town — and still, nothing compared to tomorrow.

Decidedly, **no session-hopping** this year: the center is just too large for that. Tough choices had to be made, and I finally opted for:

- Kick off the day with Presentation Skills Workshop, as discussed in my first post,
- then subsequently decide to rework my slides instead of attending the Statipedia session about collaboration between governmental statistical agencies — which I was much curious about, but the guilt was just too strong đź™‚
- and to finish the working day, attend the much attractive Mathematics, Statistics, and Political Science Introductory Overview Lecture by Andrew Gelman.

The two-hours long **Presentation Skills Workshop** was just **brilliant**! The four speakers, Dick De Veaux, Scott Berry, Chris Nachtshelm, and William Li, shared to very useful tips and tricks that were delivered in a powerfully entertaining way which kind of sweetened the fact that they sometimes hit a bit too close to home! Dick De Veaux, a former performer, gave a hilarious but very accurate **counter-example talk** — based on real examples, however sad that may be. Check back soon here for his slides. Scott Berry transmitted his passion for **passion**, stressed the importance of respecting the audience by giving them **worth their time**, and how talks can be the most important part of a statistician’s job. Chris Nachtshelm came very up close and personal with the audience, **going down in the pit**, walking around through the room, all the while bringing illuminating considerations on the **fluctuating level of technical depth** one should aim at in a talk. And, in a remarkable show of encompassing all cases, William Li aimed at specific advices for **non-native English speakers** (which I appreciate first hand), focusing on cultural differences, calibrating for an American audience (key hint: talk sports, and look straight in the eyes). A brilliant, light but at the same time **deeply useful workshop**: let’s hope that such a workshop will be held every year at JSM! Congratulations to the organizer Brian L. Wiens.

Although I was really curious to get a glimpse on how governmental statistics work: after all, when I say “I’m a statistician”, the first thing that comes to mind is indeed demographic number crunching, about which I am near clueless, I then decided to put those advices to good use and tweak my slides, and skip the 2pm-3:50pm session.

I finally crowned the day with the first of the Introductory Overview Lecture, this one by **Andrew Gelman**. Most of the readers here know his major hit blog, and hence know what to expect. Nevertheless, the **breadth of his examples** and topics came as a shock, and were a treat to the audience: I surely have never before attended a talk spanning from trenches in World War I to the **rationality of voting**! I’m sorry to have had to skip the Medallion Lecture on Stein’s Method, but my idea to broaden my horizons won the duel; to compensate, I read last night the very well done Wikipedia page on the topic, a much easier read for a jetlagged evening than the more proper book by Barbour and Chen.

Andrew’s passion and **energy** are communicative, and I was delighted to get to meet him for the first time. Remarkably, he voiced his point with the strong personality that his blog is liked for, embracing such controversial topics as **opposing social science’s Mathematical models to Statistical models** — later nuancing his stance in the catchy ** “Statistics are to Mathematics what engineering is to physics/chemistry/engineering”**, reminding me of a similar (although more gastronomically themed) statement by a former professor of mine that

*“Probabilists are to statisticians what pan-makers are to cooks and biologists to physicians”*. However and in all honesty, although I completely see the validity of this, I was less certain about his more

**head-on opposition**between “Elegant but wrong mathematical models” and “Good statistical data-based models” — it might be a cultural issue (in France statistic teams are often part of the Mathematics department — and I’m not necessarily opposed to that: there’s a fine balance to find, but that would be a forum post in itself!), or it might also be more true in Political Science than in my usual fields. I very much appreciated his call to

**the end of**, and the conclusion on the need there to avoid over-simplifying and claiming one unique cause to a behavior, but instead cleanly take the whole data into account, even if (especially if!) that means

*“pop-social science”***acknowledging that several causes are combined**.

Tomorrow is going to be a crazy day, with at some point more than half a dozen parallel talks I would like to attend at the same time — and the added concentration of presenting in the morning. I will therefore postpone until tomorrow the considerations on the social aspects of JSM, which are the other strong suit of any conference. Stay tuned for more, and please feel free to summarize in the comments the great sessions you’ve attended, or to point those that you’re looking forward to!

Great blog Julien! Sounds like a brilliant conference. Opposition between â€śElegant but wrong mathematical modelsâ€ť and â€śGood statistical data-based modelsâ€ť is definitely something that needs more discussion- this is the data analyst/mathematician clash- which side of the fence are you? Looking forward to the other posts.

I think it is quite interesting to see what happens when you fit a “wrong” model to data. It’s quite along the lines of “there are no failures, only lessons.”

Thanks, Jason! I think that, objectively, I’m more on the mathematician side than on the data-analyst’s, although I’ve been trained an engineer first. Actually, I might be more an algorithmician than a mathematician or a data analyst — but as such, I tend to really be keen on the math side, also because I am simply better at it than at data analysis, of which I lack experience!

Sir David Cox added tonight his own take on this debate (see upcoming blog post), saying, as a mathematician, that the foundations of statistics are essentially conceptual, even though their implementation is mathematical or computer-science. That’s a take that I tend to agree with more than with Andrew’s — and it would somewhat put the Statistics being to Math what … Physics are to Math too!

Michael Jordan this morning also compared Stats with Computer Science (same, see upcoming blog post, I’m finishing it right now), how the two split in the 40s-50s to start getting closer again now.

This JSM is bringing quite some fuel to those thoughts! If you have any other, please chip in đź™‚