Efficiency fit

26.2.2. Efficiency fit#

Simultaneous least-squares fit to the D0-mass distributions of D+D0(Kπ+ππ+)π+ decay candidates passing and failing the kaon-momentum requirement p(K)>1 GeV/c, to determine the efficiency of the requirement on both signal and background decays.

The input data is provided by the ROOT ntuple ntp contained in the file example-data/momentum-scale.root. The tree contains branches corresponding to the 4-momenta of the final state particles (K_P, pi1_p, pi2_p, pi3_p), a branch with the D0 mass (Dz_M), a branch with the difference between the D+ and D0 masses (DM).

Two histograms of the D0 mass, each consisting of 150 bins in the range 1.8-1.95 GeV/c2, are filled from the data. The first with candidates passing the kaon-momentum requirement (hpass), the second with candidates failing the requirement (hfail).

The least-squares are computed as

LS(θpass,θfail)=ihpass(ninipass(θpass)σi)2+ihfail(ninifail(θfail)σi)2,

where the index i runs over the histogram bins, ni is the observed number of candidates in the bin, σi is the uncertainty on ni, nipass/fail is the predicted number of candidates in the bin passing/failing the kaon-momentum requirement, which depends on some unknown parameters identified by the vector θpass/fail.

To compute the predicted number of candidates, the fit model assumes that the D0-mass distributions of the candidates passing/failing the selection can be described by a signal component peaking around the nominal D0 mass, described by a Gaussian distribution

pdfsgn(m|μ,σ)e12(mμσ2)2,

and a background component, described by an exponential distribution

pdfbkg(m|λ)eλm.

Each PDF is normalized in the fit range 1.8-1.95 GeV/c2. The predicted numbers of candidates passing/failing the selection in bin i are estimated by evaluating the above PDFs at the center of the bin mi as

nipass(Nsgn,ϵsgn,μ,σpass,Nbkg,ϵbkg,λpass)=Nsgnϵsgnpdfsgn(mi|μ,σpass)+Nbkgϵbkgpdfbkg(mi|λpass),nifail(Nsgn,ϵsgn,μ,σfail,Nbkg,ϵbkg,λfail)=Nsgn(1ϵsgn)pdfsgn(mi|μ,σfail)+Nbkg(1ϵbkg)pdfbkg(mi|λfail),

where Nsgn/bkg is the total number of signal/background candidates and ϵsgn/bkg is the signal/background efficiency, and we have assumed that the signal/background shapes may have different widths/slopes depending on whether the candidates pass/fail the kaon-momentum requirement.

The fit is developed using the following frameworks: