Fitting of Histo with Logarithmic Bins

[ANUJ_CHANDRA]

Hello ROOTers

I am trying to make a one dim histo with logarithmic bins:

short NumberofBins = 100;
float X[NumberOfBins+1];
float Xlow(1.e-1), Xhigh(1.e5);
float BinSize = (log10(Xhigh) - log10(Xlow))/NumberOfBins;
for(short i = 0; i < NumberOfBins; i++)
{
X[i] = Xlowpow(10, BinSizei);
}

TH1F *h = new TH1F(“Histname”, “Title”, NumberOfBins, X);

But when I define histo with normal bins:
TH1F *h = new TH1F(“Histname”, “Title”, 10000, 0, 10000); //with the bin size = 1. and try to fit the data set with power law function. So it gives me a slope value around -2.7.

But when I tried to fit the logarithmic bin histo with the same power law function then it gives me a slope parameter value of -1.7. I don’t know how the data slope is getting modified. Please help.

Kind Regards

[moneta]

Hi,

In the second case you have many bins which are empty of with very small content. You cannot do in this case a chi2 fit, which is biased, but you should do a likelihood fit.
Try fitting with option “L”, to see if you get the same result.
In addition you could have another effect , in the first fit, due to the fact that you are computing the function at the bin center and some bin can be very large. If you observe still a difference try to use option “I” (integral of the function in the bin).

Best Regards

Lorenzo