Join GitHub today
GitHub is home to over 40 million developers working together to host and review code, manage projects, and build software together.Sign up
Remove FluxPointsDataset chi2assym option #2645
This PR removes the
It wasn't covered by a test, and not well documented (with an example and in the stats docs, explaining it's relation to normal chi2 and likelihood profiles).
My main argument why we shouldn't offer is is that is isn't standard or useful really. At https://cxc.harvard.edu/sherpa/statistics/ there's many chi2 variants, but the asymmetric one isn't one of them. I don't recall having seen the asymmetric chi2 used in papers at all. And it's not a good technique, if the likelihood is asymmetric, then the likelihood profile method which we should add soon is much better.
This is also a follow-up to #2546, removing the last case of confusing of "likelihood" vs "stat" vs "likelihood_type" names in the Gammapy codebase.
It was used here in the SED fitting tutorial: https://travis-ci.org/gammapy/gammapy/jobs/619286087#L1804
I noticed that
@adonath - Given that it's difficult to define exactly how
@@ Coverage Diff @@ ## master #2645 +/- ## ========================================== + Coverage 91.55% 91.61% +0.05% ========================================== Files 141 141 Lines 15883 15867 -16 ========================================== - Hits 14542 14536 -6 + Misses 1341 1331 -10