Re: Minuit & large # of params (once again)

Rene Brun (Rene.Brun@cern.ch)
Thu, 12 Nov 1998 09:15:21 +0100

mikula mali wrote:
>
> This issues has been somewhat discussed several weeks ago. I need to use
> a large number of params in Minuit (~80 or so) and, as Rene explained in
> previous postings, I use the following constructor:
>
> TMinuit *aMinuit = new TMinuit(95);
>
> SEEK & SIMPLEX calls work ok, however all the Minuit functions that
> estimate parameter errors (such as MIGRAD, MINOS) dont work correctly
> for parameters 26 and higher (i.e. parabollic errors & correlations are
> correctly calculated for the first 25 parameters, but not for the rest
> of it -- somehow Minuit fixes err(p)=sqrt(2) for P>25). Is this a bug or
> I'm doing sth wrong?
>

Hi Mikula,
I have found the case of two internal static arrays in TMinuit
with a maximum dimension of 50. This probably explains your problem.
I have extended these static arrays to a maximum of 100.
These static arrays should be changed to dynamic arrays. To be done.

Rene Brun