idea for GSoC: an R package for fitting Bayesian Hierarchical Models

Above all,
tnx Ben for taking time to read about my proposal!

2008/3/24, Ben Bolker <bolker at ufl.edu>:

Antonio, Fabio Di Narzo <antonio.fabio <at> gmail.com> writes:

 >
 > I've put online a temp web page with some more info (and sources):
 >
 > http://antonio.fabio.googlepages.com/rgs%3Athergibbssampler
 >
 > Bests,
 > Antonio.
 >

  Have you seen Jouni Kerman's Umacs package?  It sounds similar
 in spirit.

Ya. But speeds are rather different.
I admittely missed a comparison with Umacs in my short demo.
However, from some early experiments (I'm doing while I'm writing), as
I suspected, my approach results being many times faster than Umacs,
even if one doesn't specify samplers as C code. Things goes even
better for my demo implementation if one tries to plug in samplers
specified as pure C code, which would further eliminate a lot of
memory allocations/deallocations behind those "rnorm()".

My aim is to obtain something which achieves decent speed, compared
with JAGS. I mean, I can easily experiment new samplers by using an
interpreted language, but if at the end I obtain something which is
*many* times slower than JAGS (which is moreover much more robust and
easier to work with), the whole stuff results being of little pratical
interest.

More: how can one really experiment a new custom sampler if doing some
thousands iterations takes forever, so that checking your sampler
pratical behaviour is a pain (I speak about my personal experience)?
That's why I want to always keep attention on speed, and give the
possibility to the user to either use R or C code at his choice, with
the ability to modify model node values in place, without unneeded
'malloc's. Ya, I would abandon pure functional style...

I will try to add a reproducible benchmark comparing my demo
implementation with JAGS and Umacs. However, I see that the problem
here is still finding someone interested in it at all.

  Something I would love to see done (not that I have the time
 and energy to supervise someone to do it right now) would be
 an R (or Python/etc.: R wouldn't necessarily be the best tool)
 to translate lmer/nlme syntax (Wilkinson-Rogers with extensions
 for specifying random factors, correlation structures, etc.)
 into a BUGS file.  It strikes me that it would be a really nice
 way to bridge the gap between what mixed-model code can do
 and what requires BUGS/MCMC.  Such models could also serve as
 (1) a way to cross-check the results of mixed model code;
 (2) a way to get started in relaxing the assumptions of mixed
 models (e.g. allowing for non-normal random effects distributions).

That sounds interesting. However, I currenlty don't have enough
know-how to work at something like it now.

Antonio.

  Ben Bolker

 ______________________________________________
 R-devel at r-project.org mailing list
 https://stat.ethz.ch/mailman/listinfo/r-devel