Scipy vs. Minuit errors

Question

Scipy vs. Minuit errors

MickaelRigault opened this issue 9 years ago · 2 comments

Is there an error difference. I should test that with fake data. See the Pull of the recovered value

Answer 1 · 2016-05-11T17:46:05.000Z

Based on MCMC minuit looks correct. I need to check scipy

Answer 2 · 2016-11-08T19:16:14.000Z

The minuit errors are estimated for the chi2 such that chi2+1 means 1 sigma.
If your errors are wrong (if the chi2/dof is far from 1), the errors on the derived parameters will be wrong