Difference between fit panel and programmatically fit

Newbie
1

Please read tips for efficient and successful posting and posting code

ROOT Version: 6.24.06
Platform: macOS Monterey 12.2.1

I have a histogram of several peaks but only want to fit one of the peaks. For that specific peak, I know to a good degree (but not as exact as a fit would yield) the parameters needed for a Gaussian fit. When use in a program like this

gtest = ROOT.TF1('gtest','gaus(0)')
gtest.SetParameters(height, mean, sigma)
hist.Fit(gtest)

I get

image
image1200×886 79.6 KB

which is not very good given that I am trying to fit the middle peak. However, when I try to fit with the Fit Panel, using the same parameters in the “Set Parameters…” box of Fit Panel, I get a much better fit
image
image1200×886 80.4 KB

Why do I get such different results even though the fit function (“gaus”, in this case) and the fit parameters are the same?

2

Probably you have a range defined in the fit panel. In any case, see how to fit within a range:
https://root.cern/doc/master/classTH1.html#a7e7d34c91d5ebab4fc9bba3ca47dabdd

3

Hi @dastudillo,

That was one of the possibilities I thought of, but I think it’s unlikely. When using the Fit Panel, I made sure to leave the range slider alone. So, does the Fit Panel automatically change the range?

Also, I am aware that it is possible to fit a peak given a range. I am trying to avoid that option and just use the height, mean, and sigma parameters instead. I have many histograms where a peak of interest can shift around horizontally. So, I want to write something that can fit the peak if given the fit parameters without a range.

4

You need:
hist.Fit(gtest, "", "", xmin, xmax)
e.g.:
hist.Fit(gtest, "", "", mean - 3. * sigma, mean + 3. * sigma)

5

@Wile_E_Coyote,

This works for most of my cases. Thank you! Any particular reason why you pick 3.? Is that what the Fit Panel does?

6

You can set whatever “xmin” and “xmax” you want.

7

Hi,
If you want to set initial parameters when fitting a Gaussian function (“gaus”), you should use the option “B”. I think this is the reason you observe such difference when using the FitPanel.

gtest = ROOT.TF1('gtest','gaus')
gtest.SetParameters(height, mean, sigma)
hist.Fit(gtest,'B')

Lorenzo

8

@moneta,

Thank you!!