TMinuit MIGRAD - Likelihood minimization

ROOT
1

Hello :slight_smile:
I am trying to fit energy of the particle by means of LogLikelihood minimization. I use TMinuit - MIGRAD for this purpose. However, if I run my code several times I obtain totally different energies - seems that likelihood function converges in wide range of energies. Do you have some suggestions, please? What can cause such instability in minimization?

Thank you for reply in advance.

2

Hi,
It is possible that some of the parameters are very correlated so the solution is not anymore a single point.
If you start with different initial points and you get different solutions, this can be caused also by the presence of several local minima.
I would suggest to check for the returned correlation and the status of the computed Hessian.

Lorenzo

1 Like
3

Thanks a lot for your suggestion.

4

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