Re: [R] ML optimization question--unidimensional unfolding scaling

From: Spencer Graves <spencer.graves_at_pdf.com>
Date: Wed 12 Oct 2005 - 13:11:30 EST

          There may be a few problems where ML (or more generally Bayes) fails to give sensible answers, but they are relatively rare.

          What is your likelihood? How many parameters are you trying to estimate?

          Are you using constrained or unconstrained optimization? If constrained, I suggest you remove the constraints by appropriate transformation. When considering alternative transformations, I consider (a) what makes physical sense, and (b) which transformation produces a log likelihood that is more close to being parabolic.

          Hou are you calling "optim"? Have you tried all "SANN" as well as "Nelder-Mead", "BFGS", and "CG"? If you are using constrained optimization, I suggest you move the constraints to Inf by appropriate transformation and use the other methods, as I just suggested.

          If you would still like more suggestions from this group, please provide more detail -- but as tersely as possible. The posting guide is, I believe, quite useful (www.R-project.org/posting-guide.html).

          spencer graves

Peter Muhlberger wrote:

> I'm trying to put together an R routine to conduct unidimensional unfolding
> scaling analysis using maximum likelihood. My problem is that ML
> optimization will get stuck at latent scale points that are far from
> optimal. The point optimizes on one of the observed variables but not
> others and for ML to move away from this 'local optimum', it has to move in
> a direction in which the likelihood is decreasing, which it won't.
>
> It's not hard to know where to look for a more optimal value--it'll be just
> on the other side of the mean of a curve. So, I can find better values, but
> these values need to be fed back into ML for continued optimization.
> Problem is, optim or nlm don't allow me to feed them new values for
> parameters and in any event ML will likely choke w/ parameters jumping
> around.
>
> One solution I've thought of is to restart optim or nlm w/ the new values
> whenever a point jumps. Is there any good way to get optim or nlm to
> prematurely terminate, return control to the calling program, while
> retaining a copy of the estimates?
>
> Perhaps ML isn't the best approach for this kind of problem. Suggestions
> welcome!
>
> Cheers,
> Peter
>
> ______________________________________________
> R-help@stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html

-- 
Spencer Graves, PhD
Senior Development Engineer
PDF Solutions, Inc.
333 West San Carlos Street Suite 700
San Jose, CA 95110, USA

spencer.graves@pdf.com
www.pdf.com <http://www.pdf.com>
Tel:  408-938-4420
Fax: 408-280-7915

______________________________________________
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Received on Wed Oct 12 13:16:35 2005

This archive was generated by hypermail 2.1.8 : Fri 03 Mar 2006 - 03:40:41 EST