X writes to remind us of the Bayesian computation conference:

– BayesComp 2020 occurs on 7-10 January 2020 in Gainesville, Florida, USA

– Registration is open with regular rates till October 14, 2019

– Deadline for submission of poster proposals is December 15, 2019

– Deadline for travel support applications is September 20, 2019

– Sessions are posted on http://users.stat.ufl.edu/~jhobert/BayesComp2020/Conf_Website/#programme

– There are four free tutorials on January 7, 2020, on Stan, NIMBLE, SAS, and AutoStat

SAS, huh?

“SAS, huh?”

Why is this so surprising?

From what I see, SAS seems to be heavily used in pharmaceutical statistics, an area that seems to be very frequentist and leans conservative because of its relationship to regulatory agencies. To me at least, it’s a bit unusual to see SAS promoted as a tool for Bayesian computation, but a quick look at the SAS 9.4 user docs shows that SAS has an option for the No-U-Turn Sampler now.

I find it interesting that SAS is moving in this direction so quickly. Perhaps the SAS Institute felt that Bayesian statistics is catching on and that they need to keep up with new features in order to keep market share.

What jumped out even more for me is AutoStat, which I hadn’t heard of. That’s not so surprising, as new toolkits come out all the time. But when I search for [autostat] on Google, I get sites pricing cars. A little further down I find the AutoStat proprietary ML toolkit? Is that the one they’re going to be teaching? The site leads with

By “statistics”, they mean regression. The only hit I find for “bayesian” on their site is “naive Bayes”. It’s proprietary, closed source, and really expensive. It’s offered by the Predictive Analytics Group, whose web page lead is

Maybe it’s something different being presented at BayesComp.