Setting confidence level for parameter errors in log-likelihood fit

Dear ROOT experts,

I’ve seen in section 7.7.4 of Chapter: FittingHistograms that I can set the confidence level of the parameter errors using MinimizerOptions::SetErrorDef(errordef). For a chi-square fit, I know I should use errordef=n^2 for a confidence level of n sigma. However, for a log-likelihood fit (which is my case), I can only find in the documentation that errordef=0.5 corresponds to 1 sigma. I’d like to know what value of errordef corresponds to 2 sigma in this case.

Thank you in advance,
Lucas

Hi,

In case of a log-likelihood fit, 2 * negative_log_likelihood is distributed as a chi2 square. So, you will have always this 0.5 factor. For n sigma’s error will be 0.5 * n^2

Lorenzo

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