I just came across this problem and I’m wondering if TH1::Fit() does the right thing, or if I have to do something special. In addition I’m fitting these histograms with other histograms (they have all the same binning), so another question is if the number of fit points should be set to the number of bins in this case.

does this mean Minuit uses the fit function evaluated only at the center of each bin (no additional points)? And what would happen then if one fits histograms with different binning?

When fitting histograms with the chisquare method, two options are supported (does not matter it histogram has fixed or variable bin widths).

method1 (default). The function is evaluated at the center of each bin
method2 (when fit option “I” is specified): the integral of the function in each bin
is computed (much slower than case 1).

Note that the objective function for the fit is called by TMinuit or TFumili
and can be seen in classes TFitter or TFumili.