Reliability of fit uncertainties when cov. matrix forced pos.-def

Roofit and RooStats
1

Dear experts,

I’m working on a rather complex templated fit using HistFactory, and as I modify the fit model the covariance matrix status occasionally changes between 3 (Full accurate covariance matrix) and 2 (Full matrix, but forced positive-definite).
Are the fit uncertainties reliable when the covariance matrix is forced pos. def.? Or could this be a bad sign for the minimum found by the fit?

Thank you in advance,
Lucas

2

Hi Lucas,

Let me put in the loop our math experts: @moneta @jonas @StephanH .

Cheers,
D

3

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.