Hello all,
It happened to me when I was trying to fit some MC data in unbinned method using the Crystal Ball, that the initial value of parameters remarkably affected the result of fitting. I was wondering whether there is some particular way to decide these initial values.
Cheers,
Fangzhou
Hi Fangzhou,
[quote=“zhangfz”]t happened to me when I was trying to fit some MC data in unbinned method using the Crystal Ball, that the initial value of parameters remarkably affected the result of fitting. I was wondering whether there is some particular way to decide these initial values.
[/quote]
This issue is with almost all fitting. The answer is very dependent on what you are trying to fit. You can sometimes get around issues by first fitting different sub-regionsonly letting certain parameters float (e.g., excluding the “signal” region and letting only the background parameters float) and then fitting the whole region. Sometimes, just making gross estimates of certain parameters (e.g., using the RMS of a histogram as the starting mean of a peak) can help as well. If these don’t work for you, you’re going to have to provide more information.
Good luck,
Charles