Dear all,
I have a question regarding the error provided by the TFractionFitter. I went through the paper (https://arxiv.org/pdf/0803.2711.pdf), documentation and this forum topics, but didn’t find any solution to my problem.
To calculate the contamination (fraction of secondary particles in the data sample), I am looking at the DCA distributions and trying to get the fraction of primaries and two different types of secondaries, which are given by MC templates, in the DCA distribution for the total number of particles, which is my data histogram ( however, I am not using the real data for now, therefore this distribution is taken from the MC as well, but just as one for all charged particles).
What I see in my final results for contamination, is that the fraction of secondaries is reasonable as it is expected, but the statistical errors which are taken from GetResult() method are overestimated because they don’t agree with the level of the fluctuation of the middle values. Such an overestimation was mentioned in a paper, which says:
Also in cases where the Monte Carlo contribution to the error cannot be neglected, TfractionFitter overestimates the error by an amount that depends on the value of p1…
, but then:
In all cases, a better estimate of the error is obtained propagating the error on formula (15), using the full covariance matrix, retrieved using methods of the class TfractionFitter.
Does somebody know what kind of method is that and can it help in correcting the errors in this situation?
I attach the example plot I described below.
Thanks a lot for any help!
Zhanna
