Closed eggplantbren closed 11 years ago
define "phase transition"
It's the spike/slab situation. If the posterior has that shape then you can't move between the spike and the slab, even though it's unimodal. I don't have an intuition for when they occur in data analysis problems. They do tend to be uncommon, but anecdotally I've seen them the most in fitting sinusoidal-type models to time series data.
On Thu, Jan 24, 2013 at 3:53 PM, David W. Hogg notifications@github.comwrote:
define "phase transition"
— Reply to this email directly or view it on GitHubhttps://github.com/dfm/turnstile/issues/1#issuecomment-12634690.
Dr Brendon J. Brewer Department of Statistics, The University of Auckland, New Zealand Ph: +64 27 500 1336 Web: http://stat.auckland.ac.nz/~brewer/
I think this is an issue for exoplanets, not for the code. Closing
As seen by Fengji, multi-planet fits, at least to RV data, have phase transitions. You'd better hope you can initialise above them. If not you may never find the good fits.