Re: Negative Sigma in Gaussian Fit on Linux/egcs1.1.1

Rene Brun (Rene.Brun@cern.ch)
Mon, 15 Mar 1999 17:19:28 +0000

Wolfgang Korsch wrote:
>
> Dear ROOT team,
>
> I tried to do gaussian fit and got the following result on LINUX RH 5.2
> w/ EGCS1.1.1:
>
> FCN=3.47372 FROM MIGRAD STATUS=CONVERGED 114 CALLS 115 TOTAL
> EDM=7.11533e-06 STRATEGY= 1 ERROR MATRIX ACCURATE
> EXT PARAMETER STEP FIRST
> NO. NAME VALUE ERROR SIZE DERIVATIVE
> 1 Constant 4.57000e+01 4.66737e+00 4.66669e-03 -1.96269e-04
> 2 Mean 1.20224e+07 4.70724e+01 5.73270e+00 2.06634e-06
> 3 Sigma -4.72385e+02 1.23139e+01 1.40189e-02 3.17310e-04
>
> The data is fine and even the fit looks fine when plotted on top of the
> data. Running the same software on DU 4.0 gives the same Sigma with
> a positive sign!!
>
> I use ROOT version 2.21/07
>
> Any ideas? Thanks.
>
> Wolfgang
>
>

Hi Wolfgang,
In 2.21/08, I have introduced the constraint to be positive for "sigma".
I think your histogram has a large mean value and small rms.
Could you send me a Root file containing your histogram?

Rene Brun