Re: [ROOT] strange error message when fitting

From: Sébastien Gadrat (gadrat@clermont.in2p3.fr)
Date: Fri May 14 2004 - 14:22:43 MEST


Dear Rene, Rooters,

I put in attachment two files, the first one is histo.root which contain 
the histogramm (named "h37") I try to fit and the second is testfit.C 
which is a reduced version of the macro I use.  This macro do two 
differents fits on the same plot (which is duplicated using the clone 
method of TH1F). I put hardwarely the fit parameters which made the 
matrix to not be defined positively but in the initial macro these are 
calculated from the parameters returned by the first fit.
I made this test macro quickly so just tell me whether I have to put the 
calculation of the new parameters step in this step macro.
Thank you for your help,
Best regards,

Sebastien

Rene Brun a écrit :

>Hi Sebastien,
>
>As explained by Eddy, it looks like your parameters are correlated.
>Could you send a small file containing your histogram, plus
>a short test macro?
>
>Rene Brun
>
>On Thu, 
>13 May 2004, Sébastien Gadrat wrote:
>
>  
>
>>Hi Rooters,
>>
>>I get a strange error message when fitting some mass plots. Among 
>>information returned by Minuit after the completion of the fit, I get 
>>"ERR MATRIX NOT POS-DEF" ! I have read in Minuit documentation the need 
>>for a positive defined matrix around the minimum for a "physical" 
>>function. I use the same method to fit five different mass spectra 
>>(differents cuts are used to generate each of these mass spectra) and 
>>only one fit returned this error message. So I think the parametrization 
>>is good. According to Minuit documentation, it could also be due to a 
>>numerical inaccuracies or a non-physical region. What can I do to 
>>distinguish between these two possibilities and to get a positived 
>>defined matrix ?
>>
>>Here is the output I get for this fit (the fit function is made of three 
>>exponentials):
>>
>>FCN=153.796 FROM MIGRAD    STATUS=CONVERGED      91 CALLS          92 TOTAL
>>                    EDM=2.47125e-08    STRATEGY= 1      ERR MATRIX NOT 
>>POS-DEF
>> EXT PARAMETER                APPROXIMATE        STEP         FIRST
>> NO.   NAME      VALUE            ERROR          SIZE      DERIVATIVE
>>  1  const0       5.72969e+03   7.36832e+01   4.07529e-01  -1.22906e-06
>>  2  const1      -8.46984e-01   7.65250e-03   3.44875e-06   7.87522e-02
>>  3  const3       9.15351e-02   7.21665e-03   1.82483e-05  -4.06709e-02
>>  4  slope1       2.19718e+00   6.64370e-03   2.93137e-06  -4.12211e-02
>>  5  slope2       2.08281e+00   6.22503e-03   2.84950e-06   4.62693e-02
>>  6  slope3       5.18135e-01   1.17508e-02   3.77755e-05   7.06570e-03
>>
>>Thanks in advance,
>>Regards,
>>
>>Sebastien
>>
>>    
>>
>
>  
>








This archive was generated by hypermail 2b29 : Sun Jan 02 2005 - 05:50:08 MET