Thursday, July 26, 2018

Stan Pharmacometrics conference in Paris July 24 2018

I just got back from attending this amazing conference in Paris:

A few people were disturbed/surprised by the fact that I am linguist ("what are you doing at an pharmacometrics conference?"). I hasten to point out that two of the core developers of Stan are linguists too (Bob Carpenter and Mitzi Morris). People seem to think that all linguists do is correct other people's comma placements. However, despite my being a total outsider to the conference, the organizers were amazingly welcoming, and even allowed me to join in the speaker's dinner, and treated me like a regular guest.

Here is a quick summary of what I learnt:

1. Gelman's talk: The only thing I remember from his talk was the statement that when economists fit multiple regression models and find that one predictor's formerly significant effect was wiped out by adding another predictor, they think that the new predictor explains the old predictor. Which is pretty funny. Another funny thing was that he had absolutely no slides, and was drawing figures in the air, and apologizing for the low resolution of the figures.

 2. Bob Carpenter gave an inspiring talk on the exciting stuff that's coming in Stan:

- Higher Speeds (Stan 2.10 will be 80 times faster with a 100 cores)

- Stan book

- New functionality (e.g., tuples, multivariate normal RNG)

- Gaussian process models will soon become tractable

- Blockless Stan is coming! This will make Stan code look more like JAGS (which is great). Stan will forever remain backward compatible so old code will not break.

My conclusion was that in the next few years, things will improve a lot in terms of speed and in terms of what one can do.

3. Torsten and Stan:

- Torsten seems to be a bunch of functions to do PK/PD modeling with Stan.

- Bill Gillespie on Torsten and Stan:

- Free courses on Stan and PK/PK modeling:

4. Mitzi Morris gave a great talk on disease mapping (accident mapping in NYC) using conditional autoregressive models (CAR). The idea is simple but great: one can model the correlations between neighboring boroughs. A straightforward application is in EEG, modeling data from all electrodes simultaneously, and modeling the decreasing correlation between neighbors. This is low-hanging fruit, esp. with Stan 2.18 coming.

5. From Bob I learnt that one should never provide free consultation (I am doing that these days), because people don't value your time then. If you charge them by the hour, this sharpens their focus. But I feel guilty charging people for my time, especially in medicine, where I provide free consulting: I'm a civil servant and already get paid by the state, and I get total freedom to do whatever I like. So it seems only fair that I serve the state in some useful way (other than studying processing differences in subject vs object relative clauses, that is).

For psycholinguists, there is a lot of stuff in pharmacometrics that will be important for EEG and visual world data: Gaussian process models, PK/PD modeling, spatial+temporal modeling of a signal like EEG. These tools exist today but we are not using them. And Stan makes a lot of this possible now or very soon now.

Summary: I'm impressed.


Unknown said...

Hi, one of the Torsten developers here. Just want to say thank you for mentioning Bill and Torsten in
the post

BTW, there's a typo in that link to Bill's paper in the post.

Unknown said...

We're better than that. We correct people's misconceptions about grammar and semantics.

I've seen Andrew make these points before, so it was easier for me to digest rapid fire. The real point of Andrew's talk was that doing the toxicometric modeling required him to think about every aspect of Bayesian statistics including thinking generatively about the process (flows among body compartments with different absorptions because the data wasn't a smooth exponential---it needed to be a mixture), two forms of strong prior information (that individual-level parameters would be grouped and that pharmacokinetic parameters were well constrained by prior experiments), how to set up inference---Frederic et al. wrote GNUsim to solve this problem, and how to do posterior predictive checks to make sure that inference made sense.

Multiple Stan books in the works---mainly one for intro grad students based on the user's guide and one to put all of our end-to-end methodolog and workflow ideas into one place. They'll both be open access and group efforts. We are converting the rest of the doc to HTML and will post.

The MPI speedups are nearly embarassingly parallel across multiple cores and GPU speedups are at factors of 50 for large matrix operations and proper memory management. Those improvements are orthogonal if you can use both in a model.

I'll be doing the talk again at StanCon in Helsinki.

The thing that Torsten helps with is preconfigured models (like RStanArm) and with handling all the funky I/O formats that are in common use for dosing and measurement.

I should've qualified more carefully. I do free consulting all the time. I love working with scientists on their problems. I'm going to visit Shravan in Berlin soon to do just that! What I meant is that you shouldn't try to lowball rates when doing consulting for pay, because you want the customer to respect your time and skills. Even then, I found my advice often being ignored. But at least then you feel like you got appropriately compensated.

I still need to get my head around what they were doing with GP emulators. People are doing that in soil carbon modeling, which uses the same kind of compartment models as the PK/PD models.

Shravan Vasishth said...

Thanks, Bob, for the detailed comments. I was away for a few weeks on vacation, no computer etc., so I only now saw your response. Didn't mean to misrepresent you there. I did miss Andrew's points, but maybe it was because I missed the start of the talk (took two hours to get from the airport to the venue!).

Shravan Vasishth said...

Thanks, Unknown. I just fixed the link.