Re: [ROOT] Problems with fitting with chi-square

From: Jacek M. Holeczek (holeczek@us.edu.pl)
Date: Tue Mar 26 2002 - 12:21:13 MET


Hi,
What you observe of perfectly normal, although seldom mentioned.
In "low statistics" case the chi-square method will underestimate the
"area", while the log-likelihood will not (there are actually two ways of
defining the chi-square method - one will underestimate the "area", the
other one will overestimate it - the method used in root will always
underestimate, as you could see).
With "high statistics" any method gives "proper" results (but still the
log-likelihood gives "more proper" results).
Conclusion - always use log-likelihood if you have low statistics or a lot
of "noise" in your spectrum.
Best regards,
Jacek.



This archive was generated by hypermail 2b29 : Sat Jan 04 2003 - 23:50:47 MET