search

StingraySoftware / stingray

Anything can happen in the next half hour (including spectral timing made easy)!
https://stingray.science/stingray
MIT License
175 stars 144 forks source link

Model bounds ignored #463

Open matteolucchini1 opened 4 years ago

matteolucchini1 commented 4 years ago

Hi everyone, I'm having an issue with the modelling interface. I'm modelling the PSD of an X-ray binary trying to identify and subtract the QPOs from the broadband noise, and the issue I'm having is that when fitting the PSD the modelling bounds I'm imposing are neglected by the fit. As a result I get unphysical results and/or poorly constrained parameters (such as negative normalisations of components that should be additive). I have uploaded a tarball including one of the lightcurves I'm working on and a Jupyter notebook to replicate the issue here: https://drive.google.com/file/d/1k8D-4F4h0H98D5fIqYTxBvcEqSWbTkLi/view?usp=sharing

matteobachetti commented 2 years ago

@dhuppenkothen I think this goes with the planned re-write of the modeling interface. Probably too complicated to address before 1.0.0? (Sorry for the extremely slow progress @matteolucchini1 , but this issue is produced by an evil mixture of Stingray's infrastructure and the somewhat inconsistent behavior of scipy's fitting interface)

abigailStev commented 2 years ago

Just came here to comment on this as well. Is there a little patch we can include for v1.0, and properly overhaul it with the rest of modeling in the future?

matteobachetti commented 2 years ago

I'm not confident that this will take a little patch 😅

abigailStev commented 2 years ago

I'll put a warning in the release notes for v1.0