[SciPy-User] [OT] Bayesian vs. frequentist
Wed Feb 15 10:11:37 CST 2012
I'm another physicist and find Silva's book to be good. One of the things
that I've used is maximum entropy in trying to reconstruct magnetization
densities from neutron scattering data, rather than Fourier transforms (sad
problems with termination effects...). I'd also like to use it more in
model selection--for example, say you have a data set that you can fit to 4
gaussians, or 2--even if you get a "better" (lower chi^2), is it
significant? BIQ can be useful...
On Wed, Feb 15, 2012 at 7:37 AM, Lou Pecora <firstname.lastname@example.org> wrote:
> *From:* Daniele Nicolodi <email@example.com>
> *To:* firstname.lastname@example.org
> *Sent:* Wednesday, February 15, 2012 3:21 AM
> *Subject:* Re: [SciPy-User] [OT] Bayesian vs. frequentist
> Hello, I'll hijack this thread to ask for advice.
> I'm a physicist and, as you may expect, my education in statistics is
> mostly in Frequentists methods. However, I always had an interest in
> Bayesian methods, as those seems to solve in much more natural ways the
> problems that arise in complex data analysis.
> I recently started to read "Data Analysis, A Bayesian Tutorial" by D.S.
> Silva (currently reading chapter 4, unfortunately real work is always
> interfering) and I really like the approach and the straight forward
> manner in which the theory builds up.
> However, I feel that the Bayesian approach, is much more difficult to
> translate to practical methods I can implement, but I may be biased by
> the long term exposition to the "recipe based" Frequentist approach.
> Can someone suggest me some resources (documentation or code) where some
> practical approaches to Bayesian analysis are taught?
> Thank you. Cheers,
> SciPy-User mailing list
> I'm also a physicist and just getting into all this. Silva's book is
> good. Here are two others I found that look good and readable. I have not
> read either all the way, but they are worth examining. You should also
> (after digesting some standard Bayesian statistics) examine the newer
> latent Dirichlet methods which look pretty powerful and seem to have a
> better way to handle and generate priors. Again, I'm a novice here, but
> these look like good avenues for a scientist trying to learn Bayesian
> (1) Udo von Toussaint, "Bayesian inference in physics", REVIEWS OF MODERN
> PHYSICS, VOLUME 83, JULY–SEPTEMBER 2011
> (2) Daniela Calvetti and Erkki Somersalo, Introduction to Bayesian
> scientific computing (Springer, 2007)
> It's a good topic even if it's OT -- provided everyone remains civil. :-)
> -- Lou Pecora, my views are my own.
> SciPy-User mailing list
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the SciPy-User