search

wbalmer / backtracks

Python package to fit relative astrometry with background star motion tracks.
https://backtracks.readthedocs.io/
BSD 3-Clause "New" or "Revised" License
6 stars 2 forks source link

Introduce host star priors, based on 5D+RV Gaia statistics (incl. covariance) #37

Closed gotten closed 8 months ago

gotten commented 8 months ago

Closes #35.

Potentially could use another diagnostic plot where host star parameters are compared with the prior to make sure it is consistent.

Might crash if Gaia has no RV datapoint.

gotten commented 8 months ago

Also setting ndim=5 in dynesty is expected to crash on making the plots because of the hardcoded best_pars. It does converge as it used to.

Edit: it didn't crash because the tests do not trigger stationtrackplot.

gotten commented 8 months ago

Also closes #38 .

gotten commented 8 months ago

Yes, I was being lazy by inserting the RV prior into the multivariate PPF. It's an independent measure anyway. We can make it a separate prior that can be overridden (with another normal distr, or uniform distr). This might give a slight performance boost too because the multivariate PPF has less dimensions.

I can separate out the RV prior but I don't know the best way of implementing an override.