1: \section{The Maximum-Likelihood Fit}
2: \label{sec:themlfit}
3: We use an unbinned extended-maximum-likelihood fit to extract
4: the $\rho^{0}\KS$ event yield. In view of a future analysis of time-dependent CP asymmetry, the events in this sample are flavour-tagged as $B^0$ or $\Bzb$ with the method described in \cite{bib:BabarS2b}. Four flavour tagging categories and an ``untagged'' category are defined each having a different expected purity for the signal.
5: The likelihood function for the $N_\cat$ candidates in flavour tagging category $k$ is
6: \begin{equation}
7: \label{eq:pdfsum}
8: {\cal L}_k = e^{-N^{\prime}_\cat}\!\prod_{i=1}^{N_\cat}
9: \bigg\{ N_{S}\epsilon_\cat\left[
10: (1-f^\cat_\textrm{MR}){\cal P}_{i,\cat}^{S-\textrm{CR}} +
11: f^\cat_\textrm{MR}{\cal P}_{i,\cat}^{S-\textrm{MR}}
12: \right]
13: + N_{C,\cat} {\cal P}_{i,\cat}^{C}
14: + \sum_{j=1}^{n_B} N_{B,j} \epsilon_{j,\cat}{\cal P}^{\B}_{ij, \cat}
15: \bigg\}
16: \end{equation}
17: where $N^{\prime}_\cat$ is the sum of the yields of all components, signal and backgrounds,
18: tagged in category $\cat$,
19: $N_S$ is the number of
20: $\rho^{0}\KS$ signal events in the sample, $\epsilon_\cat$ is the
21: fraction of signal events tagged in category $\cat$, $f^\cat_\textrm{MR}$
22: is the fraction of mis-reconstructed signal events in tagging category
23: $\cat$, $N_{C,\cat}$
24: is the number of continuum background events that are tagged in
25: category~$\cat$, and $N_{B,j}\epsilon_{j,\cat}$ is the number of
26: $B$-background events of class $j$ that are tagged in category~$\cat$.
27: The \B-background event yields are fixed in the default fit.
28: The total likelihood
29: ${\cal L}$ is the product of the likelihoods for each tagging category.
30:
31: The probability density functions (PDFs) ${\cal P}_{\cat}^{S-\textrm{CR}}$,
32: ${\cal P}_{\cat}^{S-\textrm{MR}}$,
33: ${\cal P}_{\cat}^{C}$ and ${\cal P}^{\B}_{j, \cat}$, for correctly reconstructed
34: signal, mis-reconstructed signal,
35: continuum background and $B$-background class $j$ (see Table \ref{efficiencyTable}), respectively,
36: are each the product of the PDFs of four discriminating variables.
37: Each signal and background PDF is thus given by:
38: $ {\cal P}_\cat =
39: {\cal P}(\mes)\cdot
40: {\cal P}(\de) \cdot
41: {\cal P}_\cat(\NN) \cdot
42: {\cal P}(|\cos(\theta_{\pi^+})|)$
43: where \mes, \de, \NN and $|\cos(\theta_{\pi^+})|$ are the variables described in section 2.
44:
45:
46: The $\mes$, $\de$, \NN, $|\cos{(\theta_{\pi^+})}|$ PDFs
47: for signal and $B$ background are taken from the simulation %except for
48:
49: There are 15 free parameters in the default fit, including the yields of continuum events in each of the tag categories and the signal yield.
50: Nine of the free parameters are used to describe the shape of the
51: continuum background - third order polynomials to describe the NN shape and
52: $|\cos(\theta_{\pi^+})|$, a second order polynomial to describe the $\Delta{E}$
53: distribution and for $m_{ES}$ a function describing phase space with a single
54: free parameter for its slope.
55:
56:
57:
58: