Bayesian reanalysis of the quasar dataset …

Just received a notification that the conference proceedings concerning my Bayesian reanalysis of the quasar dataset (regarding the evidence, or lack thereof, for a dipole variation in the fine structure constant) has now been published in the Mem.S.A.It journal: here.  It’s neat because this may end up being the only official publication on this work thanks to my lack of enthusiasm for dealing with the MNRAS “reviewing” process at the moment.

In a very brief astro ph skim today I noticed some good motivation for a pedagogical paper on GLMs (directed towards astronomers, of course; zoologists are already all over GLMs): Fig 1, right panel, of this paper by Mieske et al. contains the classic astronomer’s hack of binning binomial data in order to fit a straight line through the observed fractions in each bin, instead of simply fitting a logistic or probit regression model.  One day …

I also had a productive discussion with some exoplanet scientists today; and I have inspiration for a future post on the limitations of Gaussian Processes when used to model highly non-Gaussian stochastic processes.  One issue is the “nugget” problem seen in emulator studies (e.g. as addressed by GPfit), which the team didn’t seem to have encountered before.  But really the key issue concerns the nature of GPs when used as stochastic process priors in time series error models: that although the sample paths of Gaussian processes can be potentially rather complex and “non-Gaussian-looking”, if the typical (finite) sample paths of the true noise process are atypical for the GP model the resulting posterior may well suffer from the impact of this poor prior choice.

This entry was posted in Astrostatistics, Gaussian Processes. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s