juliet
juliet copied to clipboard
Raise chronological period exception more often
Hi! Not an issue, but perhaps an enhancement to make fitting RV data for multi-planet systems more user-friendly. The docs specify that for multi-planet systems, the planetary periods must be in 'chronological order'. So lets say I have a two-planet system with planet b having a 10-day period and planet c a 30-day period. If I try to fit RV data for this system, fixing the periods to those 'literature' values, and choose planet c to be p1, I get the following error:
Exception: INPUT ERROR: planetary periods in the priors are not ordered in chronological order. Planet p2 has a period of 9.0 days, while planet p1 has a period of 32.939614 days (P_p2<P_p1).
which is great and self-explanatory.
Instead, if I fix the period of planet c (p1) but put a prior on the period of planet b (p2) this exception is no longer raised - even when the prior constrains the period to always be shorter than 30-days (e.g. U(9,11) ). In this case, dataset.fit
will run without error, but finish in one iteration and return nonsense, e.g., this printed output:
1it [00:00, 63.27it/s, +300 | bound: 0 | nc: 1 | ncall: 301 | eff(%): 100.000 | loglstar: -inf < -inf < inf | logz: -inf +/- nan | dlogz: 0.000 > 0.309]
In situations like this, I think it would be helpful to raise the same exception as above.
Hi @radicamc,
Thanks for opening this issue! I see your point, and I think you are right --- at least a warning should be raised indeed based on the support of the prior if it is disjoint with the others (including fixed periods).
Cheers, Néstor