Re: gMinuit problem ..

Art,

Your analysis is right.

Rene Brun

Arthur E. Snyder wrote:
> Thanks, Rene. Something similar must be happening with our slightly more
> complex fit where sometimes gMinuit apprears and other times not.
>
> I would guess that if we move the fitting function to pre-defined
> function rather then a string in the TF1 function this would fix the
> "problem" as root would no longer be able to see if the function was
> simple or not and would always have to use Minuit.
>
> Does that sound right?
>
> -Art
>
> A.E. Snyder, Group EC \!c*p?/
> SLAC Mail Stop #95 ((. .))
> Box 4349 |
> Stanford, Ca, USA, 94309 '\|/`
> e-mail:snyder_at_slac.stanford.edu o
> phone:650-926-2701 _
> http://www.slac.stanford.edu/~snyder BaBar
> FAX:650-926-2657 Collaboration
>
>
>
> On Sat, 28 Feb 2009, Rene Brun wrote:
>
>> Art,
>>
>> When fitting with simple functions like polynomials, TMinuit is not
>> used, only the linear fitter (TLinearFitter). The linear fitter is
>> much faster than Minuit. You can force fitting of a polynomial with
>> Minuit by specifying the fit option "L", see documentation of TH1::Fit
>>
>> Rene Brun
>>
>> Arthur E. Snyder wrote:
>>>
>>> Rene et al.,
>>>
>>> We are seeing strange behavior with gMinuit. It seems to come and
>>> go. I expect it to be there after I do it fit.
>>>
>>> An example:
>>>
>>> I have a histogram (it doesn't matter what it is).
>>>
>>> root [26] h->Fit("pol4")
>>> Fitting results:
>>> Parameters:
>>> NO. VALUE ERROR
>>> 0 2.625170e+03 1.191186e+02
>>> 1 -1.565262e+04 1.055977e+03
>>> 2 3.861477e+04 3.327115e+03
>>> 3 -4.315887e+04 4.446220e+03
>>> 4 1.804345e+04 2.140152e+03
>>> (Int_t)0
>>> root [27] gMinuit
>>> (class TMinuit*)0x0
>>> root [28]
>>>
>>> -->gMinuit is MIA
>>>
>>> root [28] h->Fit("gaus")
>>> FCN=687.968 FROM MIGRAD STATUS=CONVERGED 3927 CALLS
>>> 3928 TOTAL
>>> EDM=5.19353e-06 STRATEGY= 1 ERR MATRIX
>>> NOT POS-DEF
>>> EXT PARAMETER APPROXIMATE STEP FIRST
>>> NO. NAME VALUE ERROR SIZE DERIVATIVE
>>> 1 Constant 1.93812e+09 1.54608e+08 9.24169e+02
>>> 2.24749e-10
>>> 2 Mean -8.98813e+00 2.37909e-02 2.18708e-05
>>> 1.46725e+00
>>> 3 Sigma 1.67356e+00 3.02321e-02 2.53915e-03** at
>>> limit **
>>> (Int_t)0
>>> root [29] gMinuit
>>> (class TMinuit*)0x9486650
>>>
>>> -->gMinuit is there!
>>>
>>> In a more complex sitution where we construct a simple TF1 in a
>>> macro, gMinuit is sometimes there and sometimes not -- seemingly w/o
>>> reason --
>>>
>>> -Art S.
>>>
>>>
>>>
>>> A.E. Snyder, Group EC \!c*p?/
>>> SLAC Mail Stop #95 ((. .))
>>> Box 4349 |
>>> Stanford, Ca, USA, 94309 '\|/`
>>> e-mail:snyder_at_slac.stanford.edu o
>>> phone:650-926-2701 _
>>> http://www.slac.stanford.edu/~snyder BaBar
>>> FAX:650-926-2657 Collaboration
>>>
>>>
>>>
>>
>>
>
Received on Mon Mar 02 2009 - 07:00:28 CET