Re: Fit multiple sub-ranges with TFractionFitter

  Dear Frank,

> At present, I'm afraid this functionality is not present in the
> TFractionFitter. But before answering in more detail, could you
> perhaps clarify a bit the precise model you're assuming? Are you
> assuming that the fractions of your 3 MC components are the same for
> all 5 eta ranges? Or that the "true" pt distributions are the same for
> all eta ranges?

For each data Pt distribution in a given eta bin I have the corresponding MC fractions, therefore the MC components are not the same for all the 5 eta ranges

To complicate a bit more the procedure I split the samples (both data and MCs) in positive and negative charged tracks

> Concerning your specific question of restricting the fit to multiple
> ranges: if these fit ranges are the same for all eta ranges, wouldn't
> you achieve this by creating a 2D histogram rather than a 1D one?

At first I thought TFractionFitter would work only on TH1 but by a further investigation it looks like it should work with TH2 as well

However I guess this is not enough: the point in fact is that when looking at higher eta tracks the corresponding Pt range reduces quite a lot Consequently the low stats bins are different in the different eta bins but the SetRangeX(Y) methods only allow to restrict the fit to a box

> If that is not sufficient, the only mechanism I could think of is to
> provide a method that allows the user to exclude specific bins from
> the fit; but this would already require a code update.

How long would that take (not an official release but a trunk version I can access)?

>> Something I tried is to set the bin content(error) of the low stats 
>> bins to 0 but somehow TFractionFitter refills all these empty bins in 
>> order to have a more or less
>> continuous distribution

Is this feature something you expect or an indication I am using TFractionFitter in a wrong way?

Regards,
s.