Closed hartytp closed 9 months ago
Initial framework added. The thing I'd like to see before closing this is to have the fitters provide a significance
method which returns the fit significance (Chi2 for NormalFitter
) then add a SignificanceValidator
which compares the fit significance to a user-specified threshold value.
In the previous version of the code, fit validation was done by the fitter based on a statistical test. That turned out to be overly restrictive - it's common to fit to simplified models, which don't capture all relevant physics, so the statistical tests will often fail even for a "good" fit. We should move validation out of the main fit and into separate validator classes.