On Fri, 31 Oct 2008 04:20:45 -0700, Francogrex wrote:
> This is not technical question about lisp, but i'm trying to find a lisp
> program (if anything is written already) to do statistical analysis, I'm
> especially looking for an optimization program to find the maximum
> likelihood of a function (something like the nelder-mead simplex
> algorithm or the BFGS...); I tried to look in Koders but nothing there.
> I know it's a long shot but maybe someone has or knows. Thanks.
Here is Nelder-Mead (both plain a and grid-restrained, use the latter)
for CL:
http://prxq.wordpress.com/2006/11/05/grid-restrained-nelder-mead/
BTW, maximum likelihood violates the likelihood principle. I you want to
do your statistics right, use Bayesian analysis. CL is especially good
for the latter, since you can write very fast iterations for MCMC.
HTH,
Tamas
On Oct 31, 12:39 pm, Tamas K Papp <······@gmail.com> wrote:
> BTW, maximum likelihood violates the likelihood principle. I you want to
> do your statistics right, use Bayesian analysis. CL is especially good
> for the latter, since you can write very fast iterations for MCMC.
Hi, thanks. Yes you're right, ML is not the best, and indeed I had
used something bayesian before (MCMC) and then something in between
the two schools that I am currently using now: I have coded the
"simulated annealing" algorithm into lisp, which was honestly pretty
neat. But it goes a little slow (slowness not related to lisp or any
programming language, it's just the nature of the algorithm that it's
slow) and that's why I ask about nelder-mead because I wanted some
quick and dirty alternative.