Mind of a Markov Chain Blog

[skorsky]

The owner of the blog "Mind of a Markov Chain" says he had many
problems using PyMC, so he coded his last example in R. He is an
influential blogger, maybe someone from the development team should
contact him.

http://probabilitynotes.wordpress.com/2010/08/22/global-temperature-proxy-reconstructions-bayesian-extrapolation-of-warming-w-rjags/

[A. Flaxman]

I took a quick try at replicating his project in PyMC, but I got a little stuck, due to not knowing enough R. Any ideas? is there some smoothing hidden in the plotting code? Here's what I got: http://github.com/aflaxman/pymc_temp_proxies

--Abie

________________________________________
From: py...@googlegroups.com [py...@googlegroups.com] On Behalf Of skorsky [skor...@gmail.com]
Sent: Wednesday, August 25, 2010 2:46 AM
To: PyMC
Subject: [pymc] Mind of a Markov Chain Blog

http://probabilitynotes.wordpress.com/2010/08/22/global-temperature-proxy-reconstructions-bayesian-extrapolation-of-warming-w-rjags/

--
You received this message because you are subscribed to the Google Groups "PyMC" group.
To post to this group, send email to py...@googlegroups.com.
To unsubscribe from this group, send email to pymc+uns...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/pymc?hl=en.

[A. Flaxman]

Oops, sorry for assuming the blog author is a "he".

--Abie

________________________________________
From: A. Flaxman
Sent: Wednesday, August 25, 2010 8:47 PM
To: py...@googlegroups.com
Subject: RE: [pymc] Mind of a Markov Chain Blog

[Chris Fonnesbeck]

On Aug 25, 10:47 pm, "A. Flaxman" <a...@uw.edu> wrote:
> I took a quick try at replicating his project in PyMC, but I got a little stuck, due to not knowing enough R.  Any ideas?  is there some smoothing hidden in the plotting code?  Here's what I got:  http://github.com/aflaxman/pymc_temp_proxies

That looks fine to me. I had trouble getting convergence running it as-
is, but if I use a Metropolis sampler for beta:

mc.use_step_method(Metropolis, beta)

it seems to converge just fine.

One picky point: if you specify @observed, you do not need to add a
@stochastic decorator.

cf

[A. Flaxman]

Thanks Chris. Also, I figured out that the R code snuck a "super-smoother" into the plotting code, which was making the plots there look different from mine.

I've written things up here: http://healthyalgorithms.wordpress.com/2010/08/27/mcmc-in-python-global-temperature-reconstruction-with-pymc/

--Abie

________________________________________
From: py...@googlegroups.com [py...@googlegroups.com] On Behalf Of Chris Fonnesbeck [fonne...@gmail.com]
Sent: Thursday, August 26, 2010 8:52 AM
To: PyMC
Subject: [pymc] Re: Mind of a Markov Chain Blog

mc.use_step_method(Metropolis, beta)

cf

--