Fitting a histogram with a poisson function

ROOT
1

Hi there,

I’m trying to fit a histogram (‘value’) with a poisson function.
The following code is what I tried:

 f1 = ROOT.TF1("f1","[0]*ROOT.TMath.Power(([1]/[2]),(x/[2]))*(ROOT.TMath.Exp( ([1]/[2])))/ROOT.TMath.Gamma((x/[2])+1)", 0, 5)
        f1.SetParameters(1, 1, 1)
        values.Fit('f1', 'R')
        fit = values.GetFunction('f1')
        fit.Draw('same')

However I get the following error and I don’t know why:

Error in <TFormula::Compile>:  Bad numerical expression : "ROOT.TMath.Power(([1]/[2]),(x/[2]))"
Error in <TF1::TF1>: function: f1/[0]*ROOT.TMath.Power(([1]/[2]),(x/[2]))*(ROOT.TMath.Exp(-([1]/[2])))/ROOT.TMath.Gamma((x/[2])+1) has 0 parameters instead of 1
Unknown function: f1
Traceback (most recent call last):
  File "Ansprechmuster3.py", line 242, in <module>
    fit.Draw('same')
ReferenceError: attempt to access a null-pointer

So something is wrong with “ROOT.TMath.Power(([1]/[2]),(x/[2]))” but I really don’t have an idea.
Does anybody know what’s the problem?
Thanks in advance!

2

Try:

f1 = ROOT.TF1("f1", "[0]*TMath::Power(([1]/[2]),(x/[2]))*(TMath::Exp( ([1]/[2])))/TMath::Gamma((x/[2])+1)", 0, 5)
3

Thanks for the suggestion but while I was waiting for an answer I tried this and it worked too:

def fitfunction_poisson(x, par):
            if par[2] != 0:
                if ROOT.TMath.Gamma((x[0]/par[2])+1) != 0:
                    poisson = par[0] * ROOT.TMath.Power((par[1]/par[2]),(x[0]/par[2])) * (ROOT.TMath.Exp(-(par[1]/par[2])))/ROOT.TMath.Gamma((x[0]/par[2])+1)
                else:
                    poisson = 0
            else:
                poisson = 0

            return poisson
PoissonFit = ROOT.TF1('PoissonFit', fitfunction_poisson, 0, 5, 3)
PoissonFit.SetParameter(0, 1)
PoissonFit.SetParameter(1, 1)
PoissonFit.SetParameter(2, 1)
PoissonFit.SetParNames('par0', 'par1', 'par2')
PoissonFit.SetLineColor(4)
PoissonFit.SetLineWidth(2)
PoissonFit.SetLineStyle(2)

ROOT.gStyle.SetOptFit(0111)
ROOT.gPad.Modified()

values.Fit('PoissonFit', 'R')
PoissonFit.Draw('same')

However I’ve got a new problem now:
Despite setting SetOptFit to 0111 (like it is said in the ROOT manual) the reduced chi square isn’t shown in the statistics box (parameters 0-2 are displayed though).
Any idea why?

4

:bulb: :arrow_right: Try: gStyle->SetOptFit(0157); :mrgreen:
For some additional hints see: Only display errors and name/values of parameters in a fit :wink:

5

Worked, thank you very much!

However I have to say that the ROOT manual is very misleading because there’s no mention of any octals and it clearly says that 0111 should do the job :cry: