search

California-Planet-Search / radvel

General Toolkit for Modeling Radial Velocity Data
http://radvel.readthedocs.io
MIT License
57 stars 52 forks source link

gamma and jitter with vary=False, linear=True #339

Closed tronsgaard closed 3 years ago

tronsgaard commented 3 years ago

Hi radvel team!

I think I understand how to use Tim Brandt's trick to set vary=False, linear=True and construct the gamma posterior from the remaining free parameters. Reading the note here, it also seems possible to do it for the jitter parameter instead. However, in the GaussianProcess tutorial notebook, both the gamma and jitter parameters are defined with vary=False, linear=True. It is not obvious (at least not to me) how to derive posteriors for both gamma and jitter at the same time? The problem seems underdetermined - has anyone tried it, or is this by any chance a mistake in the Gaussian Process tutorial?

Thanks, René

bjfultn commented 3 years ago

This is an error. The linearization trick is only implemented for the gamma terms. Will be fixed in the next release.