Re: Using Minuit with TGraph2DErrors class

Could you send the shortest possible RUNNING script reproducing this problem?

Rene Brun

On Mon, 17 Jul 2006, andreyk1_at_post.tau.ac.il wrote:

>
>
> Dear rooters,
>
> I'm trying to fit some complicated function using Minuit package and
> TGraph2DErrors class. The choise for Tgraph2DErrors is due to fitted function:
> it is a function of two variables (x,y).
>
> It doesn't work properly, for exmaple, if I take a simple trial function, it
> gives a very good fit, but if I switch to another function, more compicated it
> fails. The result of fitting procedure is:
>
>
> ******************************************************************************
>
> FCN=0.756195 FROM MIGRAD STATUS=FAILED 263 CALLS 264 TOTAL
> EDM=nan STRATEGY= 1 ERROR MATRIX UNCERTAINTY 100.0 per
> cent
> EXT PARAMETER APPROXIMATE STEP FIRST
> NO. NAME VALUE ERROR SIZE DERIVATIVE
> 1 MU0 1.00000e+00 nan nan nan
> 2 Omega_0 1.00000e+00 1.41421e+00 nan 0.00000e+00
> 3 gam 1.00000e+00 1.41421e+00 nan 0.00000e+00
> 4 A1 1.00000e+00 1.41421e+00 nan 0.00000e+00
> 5 sss 1.00000e+00 1.41421e+00 nan 0.00000e+00
>
> ****************************************************************************
>
> The parameters are unchanged (unity). But there should be an answer, because my
> function indeed has a minimum (from some assumptions). If I change it to only
> one parameter it is converged bu still gives unusefull answer (the initial
> parameter is remain unchanged):
>
> ****************************************************************************
>
> FCN=0.756195 FROM MIGRAD STATUS=CONVERGED 34 CALLS 35 TOTAL
> EDM=nan STRATEGY= 1 ERR MATRIX NOT POS-DEF
> EXT PARAMETER APPROXIMATE STEP FIRST
> NO. NAME VALUE ERROR SIZE DERIVATIVE
> 1 MU0 1.00000e+00 nan 3.24730e+08 nan
>
> ****************************************************************************
>
>
> Please, may be somebody can help me? Is there another method to fit a
> two-variable function?
>
>
> Thanks,
> Andrew
>
>
> ----------------------------------------------------------------
> This message was sent using IMP, the Internet Messaging Program.
>
>
Received on Mon Jul 17 2006 - 14:41:54 MEST