INLA (Integrated Nested LaPlace Approximations) for Bayesian inference

[topkatz]

Hi!

Although it's been around for about five years, I've just recently heard about INLA (Integrated Nested LaPlace Approximations), an alternative approach to Bayesian inference for latent Gaussian models that is supposed to be much faster than the minutes and hours typically spent on MCMC simulations (http://www.informatik.uni-ulm.de/ni/staff/HKestler/Reisensburg2009/talks/held.pdf).

SAS has poured a lot of effort into Bayesian analyses in the last few development cycles, and I wonder whether INLA is anywhere on their radar screens?  As you'd expect, it's already available in R (http://www.r-inla.org/).

Is INLA in the works at SAS?  If not, has it ever been on the SASware ballot?  Is there any interest in getting it into SAS?

Thanks!

--  TMK  --

[Rick_SAS · Accepted solution]

Hi topkatz,

The short answer is that it is "on the radar screen." In fact, some SAS developers are attending a seminar at a nearby univerisity on this very topic to learn more about it and to talk with researchers in this area.

When I searched the SASWare ballot, I did not see this topic, so I encourage you to add it and to "talk it up" among your colleagues. The ballot is at https://communities.sas.com/community/support-communities/ballot   Although SAS developers have their own interests and regularly attend professional meetings to learn about new research areas, we always appreciate suggestions from SAS customers about what is important to them.