hep-ex0703008/AnalysisMethod.tex
1: \section{ANALYSIS METHOD}
2: \label{sec:Analysis}
3: 
4: The $U$ and $I$ coefficients and the $\Btopipipi$ event yield are
5: determined by a maximum-likelihood fit of the signal and background model to the 
6: selected candidate events. Kinematic and event shape variables 
7: exploiting the characteristic properties of the events are used 
8: in the fit to discriminate signal from background. 
9: 
10: \subsection{EVENT SELECTION AND BACKGROUND SUPPRESSION}
11: \label{subsec:selection}
12: 
13: We reconstruct $\Btopipipi$ candidates from pairs of 
14: oppositely-charged tracks 
15: and a $\pi^0\to\gamma\gamma$ candidate.  In order to ensure that all events are within 
16: the Dalitz plot boundary, we constrain the three-pion invariant mass to the $B$ mass after final selections have been made.  
17: The largest source of background is from continuum $\epem\to q\overline{q}$ production.  
18: %
19: We use information from the tracking system, EMC, and DIRC to 
20: remove tracks for which the PID is consistent with the electron, kaon, 
21: or proton hypotheses. In addition, we require that at least one track 
22: has a signature in the IFR that is inconsistent with the muon 
23: hypothesis.  This selection retains 92\% of signal events while rejecting 
24: 42\% of continuum background events.  
25: %
26: The $\pi^0$ candidate mass $m(\gamma\gamma)$ must satisfy $0.11<m(\gamma\gamma)<0.16\gevcc$, 
27: where each photon, $\gamma$,  is required to have an energy greater than $50\mev$
28: in the laboratory frame (LAB) and to exhibit a lateral profile of energy 
29: deposition in the EMC consistent with an electromagnetic shower.
30: 
31: A $B$-meson candidate is characterized kinematically by the beam-energy substituted 
32: mass 
33: $\mes=\sqrt{{(E^{\rm cm}_{\rm beam})^2}-(p_B^{\rm cm})^2}$ 
34: and energy difference $\de = E_B^*-\half\sqrt{s}$, 
35: where $(E_B,\pvec_B)$ and $(E_0,\pvec_0)$ are the four-vectors
36: of the $B$-candidate and the initial electron-positron systems
37: respectively. The asterisk denotes the center-of-mass (CM) frame
38: and $s$ is the square of the CM energy.
39: We require $5.272 < \mes <5.288\gevcc$, which retains $81\%$
40: of the signal and $8\%$ of the continuum background events. 
41: The $\de$ resolution 
42: exhibits a dependence on the $\pi^0$ energy and therefore varies 
43: across the Dalitz plot. To avoid bias in the Dalitz plot, we introduce
44: the transformed quantity $\deprime=(2\de - \demax - \demin)/(\demax - \demin)$,
45: with $\deminmax(\mpm)=c_{\pm}-\left(c_{\pm}\mp\bar c\right)(\mpm/\mpmMax)^2$,
46: where $\mpm=\sqrt{s_0}$ is strongly correlated with the energy of the $\piz$. 
47: We use the values
48: $\bar c = 0.045\gev$, $c_{-} = -0.140\gev$, $c_{+} = 0.080\gev$,
49: $\mpmMax = 5.0\gev$, and require $-1<\deprime<1$. 
50: These values have been obtained from Monte Carlo simulation.
51: The requirement retains $75\%$ ($25\%$) of the signal (continuum) events.
52: 
53: Backgrounds arise primarily from random combinations of $\pi^\pm$ and $\pi^0$ 
54: candidates in continuum events.
55: Continuum events tend to  have a more ``jet-like'' structure than  
56: $B$ decays which are produced nearly at rest in the CM system.  
57: To enhance discrimination between signal and continuum, we 
58: use a neural network (NN)~\cite{NNo} to combine four discriminating variables: 
59: the angles with respect to the beam axis of the $B$ momentum and $B$ thrust 
60: axis in the \FourS\ frame, and the zeroth and second order polynomials
61: $L_{0,2}$ of the energy flow about the $B$ thrust axis.  The polynomials
62: are defined by $ L_n = \sum_i {p}_i\cdot\left|\cos\theta_i\right|^n$,
63: where $\theta_i$ is the angle with respect to the $B$ thrust axis of any
64: track or neutral cluster $i$, ${\bf p}_i$ is its momentum, and the sum
65: excludes the $B$ candidate.  
66: The NN is trained with off-peak data and
67: simulated signal events. The final sample of signal candidates 
68: is selected with a requirement on the NN output that retains $77\%$ ($8\%$) 
69: of the signal (continuum) events.  A total of 35444  on-peak data events pass the 
70: selection.  
71: 
72: The time difference $\deltat$ is obtained from the measured distance between 
73: the $z$ positions (along the beam direction) of the $\Bz_{\tpi}$ and 
74: $\Bz_{\rm tag}$ decay vertices, and the boost $\beta\gamma=0.56$ of 
75: the \epem\ system: $\deltat = \Delta z/\beta\gamma c$.  The $\Bz_{\rm tag}$
76: vertex is determined from the charged particles in the event not included 
77: in the  signal $B$. 
78: To determine the flavor of the $\Bz_{\rm tag}$ 
79: we use the $B$ flavor-tagging algorithm of Ref.~\cite{BabarS2b}.
80: This produces six mutually exclusive tagging categories. We improve the efficiency 
81: of the signal selection by retaining untagged events in a seventh category
82: which  contribute to the measurement of direct \CP violation. 
83: 
84: Multiple \B candidates passing the full selection occur 
85: in $16\%$ $(\rho^\pm\pi^\mp)$ and $9\%$ $(\rho^0\pi^0)$ 
86:  of $\rho(770)$ MC events. 
87: If the multiple candidates have different $\pi^0$ candidates, 
88: we choose the \B candidate with the reconstructed $\pi^0$ mass closest 
89: to the nominal $\pi^0$ mass; 
90: in the case that more than one candidate have the same $\pi^0$, we 
91: arbitrarily chose a  reconstructed 
92: \B candidates passing the selection (this occurs in $4\%$ of events).  
93: 
94: The signal efficiency determined from MC simulation is $24\%$ for 
95: $B^0 \to \rho^\pm\pi^\mp$ and $B^0 \to \rho^0\pi^0$ events, and 
96: $11\%$ for non-resonant $\Btopipipi$ events. The signal efficiency distribution on the SPD is shown in Figure \ref{fig:sigeff}.
97: \begin{figure}[tbh]
98:   \centerline{ \epsfxsize8.2cm\epsffile{effsquareall.eps}}
99:   \caption{\label{fig:sigeff}
100:         The signal efficiency distribution on the square Dalitz plot.  Note that
101: 	the plot is folded in $\thetaprime$ since the distribution 
102: 	is nearly symmetric in this variable. }
103: \end{figure}
104: 
105:  The signal events passing the event selection are a combination of correctly 
106: reconstructed (``truth-matched'', TM) events and  mis-reconstructed 
107: (``self-cross-feed'', SCF) events.
108: Of the selected signal events, $22\%$ of $B^0 \to \rho^\pm\pi^\mp$, 
109: $13\% $ of $B^0 \to \rho^0\pi^0$, and $6\%$ of non-resonant events are 
110: mis-reconstructed, according to MC.  Mis-reconstructed events occur when a track or 
111: neutral cluster from the tagging $B$ is assigned to the reconstructed signal candidate. 
112: This occurs most often for  low-momentum particles and photons; hence the mis-reconstructed
113:  events are concentrated in the corners of the standard Dalitz plot.  Since these are also the 
114: areas where the $\rho$ resonances overlap strongly, it is important to model 
115: the mis-reconstructed events correctly.  The details of the model for 
116: the distributions of mis-reconstructed 
117: events  in the Dalitz plot are described in Section \ref{sec:deltaT}.  
118: 
119: 
120: \subsection{BACKGROUND FROM OTHER {\em B} DECAYS}
121: 
122: \begin{table*}[t]
123: \begin{center}
124: \caption{ \label{tab:bbackground}
125: 	Summary of the \B-background modes taken into account for the
126: 	likelihood model. They have been grouped in 20 classes:
127:  	 charmless $B^+$ (six), charmless $B^0$ (eight), 
128: 	exclusive  charmed $B^0$ (four) and inclusive $B^0$ and
129: 	 charmed $B^+$ decays. Modes with at least two events expected
130: 	after final selection have been included.}
131: \input{bBackground.tex}
132: \vspace{-0.2cm}
133: \end{center}
134: \end{table*}
135: 
136: We use MC simulated events to study the background from other $B$ 
137: decays. More than one-hundred channels were considered in 
138: these studies, of which 29 are  included
139: in the final likelihood model.  These exclusive \B-background modes are grouped into eighteen 
140: different classes according to their kinematic and topological
141: properties: six for  charmless $\B^+$ decays, eight for charmless $B^0$ decays 
142: and four for exclusive charmed $\B^0$ decays.
143: Two additional classes account for inclusive $B^0$ and $B^+$
144: charmed decays.
145: 
146: Table \ref{tab:bbackground} summarizes the twenty background classes that are
147: used in the fit. For each mode, the expected number of selected events is
148: computed by multiplying the selection efficiency (estimated using MC
149: simulated decays) by the branching fraction, scaled to the dataset 
150: luminosity ($346\;\mathrm{fb}^{-1}$). The world average branching ratios have been
151: used for the experimentally known decay modes\cite{PDG,HFAG}. When only upper limits are
152: given, they have been translated into branching ratios including additional conservative hypotheses 
153: (e.g., 100\% longitudinal polarization for $B\to\rho\rho$ decay) if needed.
154: 
155: 
156: \subsection{THE MAXIMUM-LIKELIHOOD FIT}
157: \label{subsec:ML}
158: 
159: We perform an unbinned extended maximum-likelihood fit to extract
160: the total $\Btopipipi$ event yield, and the $U$ and $I$ coefficients
161: defined in Eqs.~(\ref{eq:firstObs})--(\ref{eq:lastObs}). 
162: The fit uses the variables  $\dt$, $\mprime$, $\thetaprime$, $\mes$, $\deprime$, and NN output 
163: to discriminate signal from background. The 
164: $\dt$ distribution is sensitive to mixing-induced \CP violation
165: but also provides additional continuum-background rejection. 
166: 
167: The selected on-resonance data sample is assumed to consist of signal, 
168: continuum-background, and \B-background components, separated by the 
169: flavor and tagging category of the tag side \B decay. 
170: The probability density function  ${\cal P}_i^\cat$ for 
171: event $i$ in tagging category $\cat$ is the sum of the probability densities 
172: of all components, namely
173: %
174: \beqn
175: \label{eq:theLikelihood}
176: 	{\cal P}_i^\cat
177: 	&\equiv& 
178: 		N_{\tpi} f^\cat_{\tpi}
179: 		\left[ 	(1-\fscfave^\cat){\cal P}_{\tpi-\TM,i}^\cat +
180: 			\fscfave^\cat{\cal P}_{\tpi-\SCF,i}^\cat 
181: 		\right] 
182: 		\nonumber\\[0.3cm]
183: 	&&
184: 		+\; N^\cat_{q\bar q}\frac{1}{2}
185: 		\left(1 + \Qtagi\Atagqq\right){\cal P}_{q\bar q,i}^\cat
186: 		\nonumber \\[0.3cm]
187: 	&&
188: 		+\; \sum_{j=1}^{N^{B^+}_{\rm class}}
189: 		N_{B^+j} f^\cat_{B^+j}
190: 		\frac{1}{2}\left(1 + \Qtagi \Atagj\right){\cal P}_{B^+,ij}^\cat
191: 		\nonumber \\[0.3cm]
192: 	&&
193: 		+\; \sum_{j=1}^{N^{B^0}_{\rm class}}
194: 		N_{B^0j} f^\cat_{B^0j}
195: 		{\cal P}_{B^0,ij}^\cat~,
196: \eeqn
197: where
198: 	$N_{\tpi}$ is the total number of $\pi^+\pi^-\pi^0$ signal events 
199: 	in the data sample;
200: %
201:  	$f^\cat_{\tpi}$ is the fraction of signal events that are 
202:        	 in tagging category $\cat$;
203: %
204: 	$\fscfave^\cat$ is the fraction of SCF events in tagging category $\cat$, 
205: 	averaged over the Dalitz plot;
206: %
207: 	${\cal P}_{\tpi-\TM,i}^\cat$ and ${\cal P}_{\tpi-\SCF,i}^\cat$
208: 	are the products of PDFs of the discriminating variables used
209: 	in tagging category $\cat$ for TM and SCF
210: 	events, respectively; 
211: %
212:  	$N^\cat_{q\bar q}$ is the number of continuum events that are 
213: 	in tagging category $\cat$;
214: %
215: 	$\Qtagi$ is the tag flavor of the event, defined to be 
216: 	$+1$ for a $\Bz_{\rm tag}$ and $-1$ for a $\Bzb_{\rm tag}$; 
217: %	
218: 	$\Atagqq$ parameterizes possible flavor tag asymmetry in continuum events; 
219: %
220: 	${\cal P}_{q\bar q,i}^\cat$ is the continuum PDF for tagging 
221: 	category $\cat$;
222: %
223: 	$N^{B^+}_{\rm class}$ ($N^{B^0}_{\rm class}$) is the number of 
224: 	charged (neutral) $B$-related background classes considered in the fit;
225: %
226: 	$N_{B^+j}$ ($N_{B^0j}$) is the number of expected events in
227: 	the charged (neutral) $B$-background class $j$;
228: %
229: 	$f^\cat_{B^+j}$ ($f^\cat_{B^0j}$) is the fraction of 
230: 	charged (neutral) $B$-background events of class $j$
231: 	that are in tagging category $\cat$;
232: %
233: 	$\Atagj$ describes a possible flavor tag asymmetry in the $B^+$ background
234: 	class $j$; 
235: %
236: 	${\cal P}_{B^+,ij}^\cat$ is the $B^+$-background PDF for tagging 
237: 	category $\cat$ and class $j$;
238: %
239: 	and ${\cal P}_{B^0,ij}^\cat$ is the neutral-$B$-background 
240: 	PDF for tagging category $\cat$ and class $j$.
241: %
242: Correlations between the flavor tag and the position in the Dalitz plot 
243:  are absorbed in tag-flavor-dependent 
244: 	Dalitz plot PDFs that are used for $B^+$ and continuum
245: 	background.
246: 
247: The PDFs ${\cal P}_{X}^{\cat}$ ($X=\{{\rm TM, SCF}, {q\overline{q}}, {B^+/B^0}\}$)
248: are the product of the four PDFs of the discriminating variables,
249: $x_1 = m_{ES}$, $x_2 = \deprime$, and $x_3 = {\rm NN~output}$, and the triplet
250: $x_4 = \{\mprime, \thetaprime, \deltat\}$:
251: \beq
252: \label{eq:likVars}
253: 	{\cal P}_{X,i(j)}^{\cat} \;\equiv\; 
254: 	\prod_{k=1}^4 P_{X,i(j)}^\cat(x_k)~,
255: \eeq
256: where $i$ is the event index and $j$ is a $B$-background class.  The extended likelihood over all tagging categories is given by
257: %
258: \beq
259: 	{\cal L} \;\equiv\;  
260: 	\prod_{\cat=1}^{7} e^{-\overline N^\cat}\,
261: 	\prod_{i=1}^{N^\cat} {\cal P}_{i}^\cat~,
262: \eeq
263: %
264: where $\overline N^\cat$ is the total number of events expected in category 
265: $\cat$. 
266: 
267: A total of 68 parameters, including the inclusive signal yield $N_{\tpi}$ and  the 26
268: $U$ and $I$ coefficients from Eq.~(\ref{eq:dt}), are varied in the fit. Most of the
269: parameters describing the continuum distributions are also free
270: in the fit.  The parameterizations of the PDFs are described below and are summarized in Tab. \ref{tab:pdfparameterization}.
271: 
272: \begin{table*}[t]
273: \begin{center}
274: \caption{ \label{tab:pdfparameterization}
275:         Summary of PDF parameterizations where G=Gaussian, PX=X-order polynomial, NP=non-parametric, and biCB=bifurcated Crystal Ball. See Section \ref{sec:deltaT} for a detailed description of the Dalitz plot parameterization for signal.}
276: \input{pdfTable.tex}
277: \vspace{-0.2cm}
278: \end{center}
279: \end{table*}
280: 
281: \subsubsection{\boldmath THE $\dt$ AND DALITZ PLOT PDFS}
282: \label{sec:deltaT}
283: 
284: 	The Dalitz plot PDFs require as input the Dalitz plot-dependent 
285: 	relative selection efficiency $\e=\e(\mprime,\thetaprime)$, 
286: 	and the SCF fraction, $\fscf=\fscf(\mprime,\thetaprime)$.
287: 	Both quantities are taken from MC simulation. 
288: 	Away from the Dalitz plot corners the efficiency is uniform, while it 
289: 	decreases when approaching the corners where one  of the 
290: 	three particles in the final state is almost at rest in the LAB frame so that the 
291: 	acceptance requirements on the particle reconstruction become 
292:         restrictive.
293: 	Combinatorial backgrounds, and hence SCF fractions, are large in
294: 	the corners
295: 	of the Dalitz plot due to the presence of soft neutral clusters 
296: 	and tracks. 
297: 
298: 	For an event~$i$, we define the time-dependent Dalitz plot PDFs
299: 	\beqn
300: 		P_{\tpi-\TM,i}^{c} &\equiv&
301: 		\varepsilon_i\,(1 - \fscfi^{c})\,\detJi\,\AmpAll~,
302: 		\\[0.3cm]
303: 		P_{\tpi-\SCF,\,i}^{c} &\equiv&
304: 		\varepsilon_i\,\fscfi^{c}\,\detJi\,\AmpAll~,
305: 	\eeqn	
306: 	where $P_{\tpi-\TM,i}$ and $P_{\tpi-\SCF,\,i}$ are normalized. The 
307: 	 normalization involves the expectation values 	
308: 	$\langle \varepsilon\,(1-\fscf)\,\detJ \,f^\kappa f^{\sigma*}\rangle$
309: 	and 
310: 	$\langle \varepsilon\,\fscf\,\detJ\, f^\kappa f^{\sigma*}\rangle$
311: 	for TM and SCF events, where the indices $\kappa$, $\sigma$ 
312: 	run over all resonances belonging to the signal model.
313: 	The expectation values are model-dependent and are 
314: 	computed with the use of MC integration over the square Dalitz plot:
315: 	\beqn
316: 	\label{eq:normAverage}
317: 		\langle \varepsilon\,(1-\fscf)\,\detJ\, f^\kappa f^{\sigma*}\rangle
318: 		\;=\; \nonumber\\
319: 		\frac{\int_0^1\int_0^1 
320: 			    \varepsilon\,(1-\fscf)\,\detJ\, f^\kappa f^{\sigma*}
321: 			\,d\mprime d\thetaprime}
322: 		       {\int_0^1\int_0^1 \varepsilon\,\detJ\, f^\kappa f^{\sigma*}
323: 			\,d\mprime d\thetaprime}~,
324: 	\eeqn
325: 	and similarly for 
326: 	$\langle \fscf \varepsilon\,\,\detJ\, f^\kappa f^{\sigma*}\rangle$,
327: 	where all quantities in the integrands are Dalitz-plot dependent.
328: 
329: 	Equation~(\ref{eq:theLikelihood}) invokes the phase 
330: 	space-averaged SCF fraction 
331: 	$\fscfave\equiv\langle\fscf\,\detJ\, f^\kappa f^{\sigma*}\rangle$. 
332: 	The PDF normalization  is decay-dynamics-dependent
333: 	and is computed iteratively. We 
334: 	determine the average SCF fractions separately for each tagging category 
335: 	from MC simulation. 
336: 	
337: 	The width of the dominant $\rho(770)$ resonance is large compared 
338: 	to the mass resolution for TM events (about $8\mevcc$ Gaussian
339: 	resolution). We  therefore neglect resolution effects in the TM 
340: 	model.	
341: 	Mis-reconstructed events	have a poor mass resolution that strongly 
342: 	varies across the Dalitz plot. These events are described in the fit by a 
343: 	two-dimensional resolution function
344: 	\beq
345: 	\label{eq:rscf}
346: 		\Rscf(\mprime_r,\thetaprime_r,\mprime_t,\thetaprime_t)~,
347: 	\eeq
348: 	which represents the probability to reconstruct at the coordinate
349: 	$(\mprime_r,\thetaprime_r)$ an event that has the true coordinate 
350: 	$(\mprime_t,\thetaprime_t)$.  This function obeys the unitary condition
351: 	\beq
352: 		\intl_0^1\intl_0^1 
353: 		\Rscf(\mprime_r,\thetaprime_r,\mprime_t,\thetaprime_t)
354: 		\,d\mprime_r d\thetaprime_r = 1,~
355: 	\eeq
356: 	and is convolved with the signal model. 
357: 	The $\Rscf$ function is obtained from MC simulation.
358: 
359: %	We use the signal model described in Section~\ref{sec:kinmeatics}. 
360: 	The dynamical information in the signal model is described in 
361: 	Section~\ref{sec:kinmeatics} and is connected with $\dt$ via 
362: 	the matrix element in Eq. (\ref{eq:dt}), which serves as the PDF.
363: 	The PDF is modified by the effects of mistagging and the limited vertex 
364: 	resolution~\cite{rhopipaper}. 
365: 	The $\deltat$ resolution function for signal and \B-background 
366: 	events is a sum of three Gaussian distributions, with parameters 
367: 	determined by a fit to fully reconstructed $\Bz$ 
368: 	decays~\cite{BabarS2b}. Since the majority of SCF events 
369: 	arise from mis-reconsructed $\pi^0$ decays which do not 
370: 	affect the vertex resolution, we use the same resolution function for
371: 	TM and SCF events. 
372: %\\[0.3cm]\noindent
373: 
374: 	The Dalitz plot- and $\dt$-dependent PDFs factorize for the 
375: 	charged-\B background modes, but not necessarily
376: 	for the $B^0$ background due to $\BzBzb$ mixing.
377: 
378:  	The charged \B-background
379: 		contribution to the likelihood~(\ref{eq:theLikelihood})
380:                 involves 
381: 		the parameter $\Atag$, multiplied by the tag flavor $\Qtag$ of 
382: 		the event. In the presence of significant ``tag-`charge'' 
383: 		correlation (represented by an effective 
384: 		flavor tag versus Dalitz coordinate correlation),
385: 		it parameterizes possible fake direct \CP violation or 
386: 		asymmetries due to detector effects in these events.
387: 		We also use separate square Dalitz plot PDFs for 
388: 	 	 $B^0$ and $\overline{B}^0$ flavor tags, and a flavor-tag-averaged PDF for 
389: 		untagged events. The PDFs are obtained from MC simulation and are 
390: 		described with the use of non-parametric functions.
391: 		The $\dt$ resolution parameters are determined by a fit to fully 
392: 		reconstructed $\Bp$ decays. For each $\Bp$-background class we obtain 
393: 		effective lifetimes from MC to account for the mis-reconstruction of the 
394: 		event that modifies the nominal $\dt$ resolution function.
395: 
396: 	The neutral-$B$ background is parameterized with PDFs that
397: 		depend on the flavor tag of the event. In the case of \CP
398: 		eigenstates, correlations between the flavor tag and the Dalitz 
399: 		coordinate are expected to be small. However, non-\CP  eigenstates,
400: 		such as $a_1^\pm\pi^\mp$, may exhibit such correlations. Both types 
401: 		of decays can have direct
402: 		and mixing-induced \CP  violation. A third type of decay
403: 		involves charged kaons (e.g. $\rho^\pm K^\mp$) 
404: 		and does not exhibit mixing-induced
405: 		\CP  violation, but usually has a strong correlation between the
406: 		flavor tag and the Dalitz plot coordinate, because 
407: 		these decays correspond to  $B$-flavor eigenstates.
408: 		The Dalitz plot PDFs are obtained from MC simulation and are 
409: 		described with the use of non-parametric functions.
410: 		For neutral-$B$ background, the signal $\dt$ resolution model 
411: 		is assumed.
412: 
413: 	The Dalitz plot
414: 		treatment of the continuum events is similar to that used
415: 		for charged-$B$ background. 
416: 		The square Dalitz plot PDF for continuum background is 
417: 		obtained from on-resonance events selected in the
418: 		$\mes$ sidebands (defined as $5.225<\mes<5.265$) 
419: 		and corrected for a 5\% feed-through
420: 		from \B decays. A large number of cross checks have been 
421: 		performed to ensure the high fidelity of the empirical shape 
422: 		parameterization. 
423: 		The continuum $\deltat$ distribution is parameterized as the sum of 
424: 		three Gaussian distributions with common mean and 
425: 		three distinct widths.  The widths scale with the estimated $\dt$ uncertainty for each event.
426: 		This yields six shape parameters that are determined by 
427: 		the fit.
428:  		The model is motivated by the observation that 
429: 		the $\dt$ average is independent of its error, and that the 
430: 		$\dt$ RMS depends linearly on the $\dt$ error.
431: 
432: \subsubsection{PARAMETERIZATION OF THE OTHER VARIABLES}
433: \label{sec:likmESanddE}
434: 
435: 	The $\mes$ distribution of TM signal events is
436: 		parameterized by a bifurcated Crystal Ball function~\cite{PDFsCB},
437: 		which is a combination of a one-sided Gaussian and 
438: 		a Crystal Ball function, given as:
439: 		\beqn
440: 		f(x) = 
441: 		\begin{cases} C e^{(x-m)^2/2s_R^2}   & \hspace{-1.0cm} \text{for $(x-m) > 0$,}
442: 		\\
443: 		        C e^{(x-m)^2/2s_L^2}&\hspace{-1.5cm} \text{for $0 > \frac{x-m}{s_L} > -A$,}
444: 		\\
445: 		       C (\frac{b}{A})^b e^{-\frac{A^2}{2}}\left(\frac{b}{A}-A-\frac{x-m}{s_L}\right)^{-b} &\hspace{-0.3cm} \text{for $\frac{x-m}{s_L} < -A$.}
446: 		\end{cases}
447: 		\eeqn
448: 		The peak position of this function, $m$, 
449: 		is determined by the fit to on-peak data while the other parameters are 
450: 		taken from signal MC. A non-parametric
451: 		function \cite{keys} is used to describe the SCF signal component.
452: 
453: 	The $\deprime$ distribution of TM events is
454: 		parameterized by a double Gaussian function, where
455: 		all five parameters depend linearly on $\mpm^2$.
456: 		The parameters of the narrow Gaussian are determined 
457: 		by the fit to data while the others are obtained from 
458: 		signal MC. 
459: 		Mis-reconstructed events are parameterized by a broad
460: 		single Gaussian function whose parameters are taken from signal MC.
461: 		
462: 	Both $\mes$ and $\deprime$ PDFs are parameterized by non-parametric
463: 		functions for all $B$-background classes.  Continuum events are
464: 		parameterized with an Argus shape function~\cite{PDFsArgus}
465: 		\beq
466: 		f(\mes)=C\frac{\mes}{m_{\rm ES}^{\rm max}}\sqrt{1-\left(\frac{\mes}{m_{\rm ES}^{\rm max}}\right)^2} e^{-\xi(1-\left(\frac{\mes}{m_{\rm ES}^{\rm max}}\right)^2)}
467: 		\eeq
468: 		 and 
469: 		a second-order polynomial in $\deprime$,  with parameters 
470: 		determined by the fit.  The value of $m_{ES}^{max}$ is 5.2886 \gevcc.
471: 
472: 	We use non-parametric functions to empirically describe the distributions 
473: 		of the NN outputs
474: 		found in the MC simulation for TM and SCF signal events, 
475: 		and for \B-background events. We distinguish tagging categories 
476: 		for TM signal events to account for differences observed in the 
477: 		shapes.
478: 	
479: 	The continuum NN distribution is parameterized by a 
480: 		third-order polynomial.
481: 		The coefficients of the polynomial are determined by the fit.
482: 		Continuum events exhibit a correlation between the Dalitz plot 
483: 		coordinate
484: 		and the inputs to the NN. 
485: 	To account for this correlation, 
486: 		we introduce a linear dependence of the polynomial coefficients
487: 		on the distance of the Dalitz plot coordinate from kinematic 
488: 		boundaries of the Dalitz plot. The parameters describing this
489: 		dependence are determined by the fit.
490: 
491: