hep-ex0608002/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 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, which are required to form a good quality vertex,
15: and a $\pi^0$ candidate.  In order to ensure that all events are within 
16: the Dalitz plot bounaries, we constrain the three-pion invariant mass to the B-mass.  
17: %
18: We use information from the tracking system, EMC, and DIRC to 
19: remove tracks for which the PID is consistent with the electron, kaon, 
20: or proton hypotheses. In addition, we require that at least one track 
21: has a signature in the IFR that is inconsistent with the muon 
22: hypothesis.
23: %
24: The $\pi^0$ candidate mass must satisfy $0.11<m(\gamma\gamma)<0.16\gevcc$, 
25: where each photon is required to have an energy greater than $50\mev$
26: in the laboratory frame (LAB) and to exhibit a lateral profile of energy 
27: deposition in the EMC consistent with an electromagnetic shower.
28: 
29: A $B$-meson candidate is characterized kinematically by the energy-substituted 
30: mass $\mes=\lbrack{(\half s+\pvec_0\cdot\pvec_B)^2/E_0^2-\pvec_B^2}\rbrack^\half$
31: and energy difference $\de = E_B^*-\half\sqrt{s}$, 
32: where $(E_B,\pvec_B)$ and $(E_0,\pvec_0)$ are the four-vectors
33: of the $B$-candidate and the initial electron-positron system,
34: respectively. The asterisk denotes the \FourS\  frame,
35: and $s$ is the square of the invariant mass of the electron-positron system.  
36: We require $5.272 < \mes <5.288\gevcc$, which retains $81\%$
37: of the signal and $8\%$ of the continuum background events. 
38: The $\de$ resolution 
39: exhibits a dependence on the $\pi^0$ energy and therefore varies 
40: across the Dalitz plot. We account for this effect by introducing
41: the transformed quantity $\deprime=(2\de - \demax - \demin)/(\demax - \demin)$,
42: with $\deminmax(\mpm)=c_{\pm}-\left(c_{\pm}\mp\bar c\right)(\mpm/\mpmMax)^2$,
43: where $\mpm$ is strongly correlated with the energy of $\piz$. 
44: We use the values
45: $\bar c = 0.045\gev$, $c_{-} = -0.140\gev$, $c_{+} = 0.080\gev$,
46: $\mpmMax = 5.0\gev$, and require $-1<\deprime<1$. 
47: These values have been obtained from Monte Carlo simulation and
48: are tuned to maximize the selection of correctly reconstructed over 
49: misreconstructed signal events. The requirement retains $75\%$ ($25\%$)
50: of the signal (continuum).
51: 
52: Backgrounds arise primarily from random combinations in continuum events.
53: To enhance discrimination between signal and continuum, we 
54: use a neural network (NN)~\cite{NNo} to combine four discriminating variables: 
55: the angles with respect to the beam axis of the $B$ momentum and $B$ thrust 
56: axis in the \FourS\ frame, and the zeroth and second order monomials
57: $L_{0,2}$ of the energy flow about the $B$ thrust axis.  The monomials
58: are defined by $ L_j = \sum_i {\bf p}_i\times\left|\cos\theta_i\right|^j$,
59: where $\theta_i$ is the angle with respect to the $B$ thrust axis of
60: track or neutral cluster $i$, ${\bf p}_i$ is its momentum, and the sum
61: excludes the $B$ candidate.  
62: The NN is trained in the signal region with off-resonance data and
63: simulated signal events. The final sample of signal candidates 
64: is selected with a requirement on the NN output that retains $77\%$ ($8\%$) 
65: of the signal (continuum).
66: 
67: The time difference $\deltat$ is obtained from the measured distance between 
68: the $z$ positions (along the beam direction) of the $\Bz_{\tpi}$ and 
69: $\Bz_{\rm tag}$ decay vertices, and the boost $\beta\gamma=0.56$ of 
70: the \epem\ system: $\deltat = \Delta z/\beta\gamma c$.
71: To determine the flavor of the $\Bz_{\rm tag}$ 
72: we use the $B$ flavor tagging algorithm of Ref.~\cite{BabarS2b}.
73: This produces six mutually exclusive tagging categories. We also 
74: retain untagged events in a seventh category to improve the efficiency 
75: of the signal selection and because these events contribute to the 
76: measurement of direct \CP violation. Events with multiple \B 
77: candidates passing the full selection occur 
78: in $16\%$ $(\rho^\pm\pi^\mp)$ and $9\%$ $(\rho^0\pi^0)$ 
79: of the cases. If the multiple candidates have different $\pi^0$'s, 
80: we choose the candidate with the reconstructed $\pi^0$ mass closest 
81: to the nominal one; in the case that both candidates have the same $\pi^0$,
82: we pick the first one.
83: 
84: The signal efficiency determined from MC simulation is $24\%$ for 
85: $B^0 \to \rho^\pm\pi^\mp$ and $B^0 \to \rho^0\pi^0$ events, and 
86: $11\%$ for non-resonant $\Btopipipi$ events. 
87: 
88: Of the selected signal events, $22\%$ of $B^0 \to \rho^\pm\pi^\mp$, 
89: $13\% $ of $B^0 \to \rho^0\pi^0$, and $6\%$ of non-resonant events are 
90: misreconstructed.  Misreconstruced events occur when a track or 
91: neutral cluster from the tagging $B$ is assigned to the reconstructed signal candidate. 
92: This occurs most often for  low-momentum tracks and photons and hence the misreconstructed events 
93: concentrate in the corners of the Dalitz plot.  Since these are also the areas where the $\rho$-mesons
94: overlap strongly, it is important to model the misreconstruced events correctly.  The details of the model
95: for misreconstructed events over the Dalitz plot is detailed in Section \ref{sec:deltaT}.
96: 
97: 
98: \subsection{BACKGROUND FROM OTHER {\em B} DECAYS}
99: 
100: \begin{table*}[t]
101: \begin{center}
102: \input{bBackground.tex}
103: \vspace{-0.2cm}
104: \caption{ \label{tab:bbackground}
105: 	Summary of the \B-background modes taken into account for the
106: 	likelihood model. They have been grouped in twenty classes:
107: 	charged charmless (six), neutral charmless (eight), 
108: 	exclusive neutral charmed (four) and inclusive neutral and
109: 	charged charmed decays. Modes with at least two events expected
110: 	after final selection have been included.}
111: \end{center}
112: \end{table*}
113: 
114: We use MC simulated events to study the background from other $B$ 
115: decays. More than a hundred channels have been considered in the
116: preliminary studies, of which twenty-nine have been finally included
117: in the likelihood model -- decays with at least two events expected after
118: selection. These exclusive \B-background modes are grouped into eighteen 
119: different classes gathering decays with similar kinematic and topological
120: properties: six for charged charmless \B-decays, eight for neutral charmless
121: \B-decays and four for exclusive neutral charmed \B-decays.
122: Two additional classes account for inclusive neutral and charged 
123: $b\to c$ decays.
124: 
125: Table \ref{tab:bbackground} summarizes the twenty background classes which are
126: used in the fit. For each mode, the expected number of selected events is
127: computed by multiplying the selection efficiency (estimated using MC
128: simulated decays) by the branching fraction scaled up to the dataset 
129: luminosity ($310\;\mathrm{fb}^{-1}$). The world average branching ratios have been
130: used for the experimentally known decay modes. When only upper limits are
131: given, they have been translated into branching ratios using all information
132: available such as additional conservative hypotheses (e.g. 100\% longitudinal
133: polarization for $B\to\rho\rho$ decay) if needed.
134: 
135: 
136: \subsection{THE MAXIMUM-LIKELIHOOD FIT}
137: \label{subsec:ML}
138: 
139: We perform an unbinned extended maximum-likelihood fit to extract
140: the inclusive $\Btopipipi$ event yield and the $U$ and $I$ coefficients
141: defined in Eqs.~(\ref{eq:firstObs})--(\ref{eq:lastObs}). 
142: The fit uses the variables $\mes$, $\deprime$, the NN output, and the 
143: Dalitz plot to discriminate signal from background. The 
144: $\dt$ measurement allows to the determination of mixing-induced \CP violation
145: and provides additional continuum-background rejection. 
146: 
147: The selected on-resonance data sample is assumed to consist of signal, 
148: continuum-background and \B-background components, separated by the 
149: flavor and tagging category of the tag side \B decay. 
150: The signal likelihood consists of the sum of a correctly 
151: reconstructed (``truth-matched'', TM) component and a misreconstructed 
152: (``self-cross-feed'', SCF) component.
153: 
154: The probability density function (PDF) ${\cal P}_i^\cat$ for an
155: event $i$ in tagging category $\cat$ is the sum of the probability densities 
156: of all components, namely
157: %
158: \beqn
159: \label{eq:theLikelihood}
160: 	{\cal P}_i^\cat
161: 	&\equiv& 
162: 		N_{\tpi} f^\cat_{\tpi}
163: 		\left[ 	(1-\fscfave^\cat){\cal P}_{\tpi-\TM,i}^\cat +
164: 			\fscfave^\cat{\cal P}_{\tpi-\SCF,i}^\cat 
165: 		\right] 
166: 		\nonumber\\[0.3cm]
167: 	&&
168: 		+\; N^\cat_{q\bar q}\frac{1}{2}
169: 		\left(1 + \Qtagi\Atagqq\right){\cal P}_{q\bar q,i}^\cat
170: 		\nonumber \\[0.3cm]
171: 	&&
172: 		+\; \sum_{j=1}^{N^{B^+}_{\rm class}}
173: 		N_{B^+j} f^\cat_{B^+j}
174: 		\frac{1}{2}\left(1 + \Qtagi \Atagj\right){\cal P}_{B^+,ij}^\cat
175: 		\nonumber \\[0.3cm]
176: 	&&
177: 		+\; \sum_{j=1}^{N^{B^0}_{\rm class}}
178: 		N_{B^0j} f^\cat_{B^0j}
179: 		{\cal P}_{B^0,ij}^\cat~,
180: \eeqn
181: where: 
182: 	$N_{\tpi}$ is the total number of $\pi^+\pi^-\pi^0$ signal events 
183: 	in the data sample;
184: %
185:  	$f^\cat_{\tpi}$ is the fraction of signal events that are 
186:        	tagged in category $\cat$;
187: %
188: 	$\fscfave^\cat$ is the fraction of SCF events in tagging category $\cat$, 
189: 	averaged over the Dalitz plot;
190: %
191: 	${\cal P}_{\tpi-\TM,i}^\cat$ and ${\cal P}_{\tpi-\SCF,i}^\cat$
192: 	are the products of PDFs of the discriminating variables used
193: 	in tagging category $\cat$ for TM and SCF
194: 	events, respectively; 
195: %
196:  	$N^\cat_{q\bar q}$ is the number of continuum events that are 
197: 	tagged in category $\cat$;
198: %
199: 	$\Qtagi$ is the tag flavor of the event, defined to be 
200: 	$+1$ for a $\Bz_{\rm tag}$ and $-1$ for a $\Bzb_{\rm tag}$; 
201: %	
202: 	$\Atagqq$ parameterizes possible tag asymmetry in continuum events; 
203: %
204: 	${\cal P}_{q\bar q,i}^\cat$ is the continuum PDF for tagging 
205: 	category $\cat$;
206: %
207: 	$N^{B^+}_{\rm class}$ ($N^{B^0}_{\rm class}$) is the number of 
208: 	charged (neutral) $B$-related background classes considered in the fit;
209: %
210: 	$N_{B^+j}$ ($N_{B^0j}$) is the number of expected events in
211: 	the charged (neutral) $B$-background class $j$;
212: %
213: 	$f^\cat_{B^+j}$ ($f^\cat_{B^0j}$) is the fraction of 
214: 	charged (neutral) $B$-background events of class $j$
215: 	that are tagged in category $\cat$;
216: %
217: 	$\Atagj$ describes a possible tag asymmetry in the charged-$B$ background
218: 	class $j$; 
219: 	correlations between the tag and the position in the Dalitz plot 
220: 	(the ``charge'') are absorbed in tag-flavor-dependent 
221: 	Dalitz plot PDFs that are used for charged-\B and continuum
222: 	background;
223: %
224: 	${\cal P}_{B^+,ij}^\cat$ is the $B^+$-background PDF for tagging 
225: 	category $\cat$ and class $j$;
226: %
227: 	finally, ${\cal P}_{B^0,ij}^\cat$ is the neutral-$B$-background 
228: 	PDF for tagging category $\cat$ and class $j$.
229: %
230: 
231: The PDFs ${\cal P}_{X}^{\cat}$ ($X=\{TM, SCF, {\rm continuum}, {\rm B-bkg}\}$)
232: are the product of the four PDFs of the discriminating variables,
233: $x_1 = m_{ES}$, $x_2 = \deprime$, $x_3 = {\rm NN output}$, and the triplet
234: $x_4 = \{\mprime, \thetaprime, \deltat\}$:
235: \beq
236: \label{eq:likVars}
237: 	{\cal P}_{X,i(j)}^{\cat} \;\equiv\; 
238: 	\prod_{k=1}^4 P_{X,i(j)}^\cat(x_k)~.
239: \eeq
240: The extended likelihood over all tagging categories is given by
241: %
242: \beq
243: 	{\cal L} \;\equiv\;  
244: 	\prod_{\cat=1}^{7} e^{-\overline N^\cat}\,
245: 	\prod_{i}^{N^\cat} {\cal P}_{i}^\cat~,
246: \eeq
247: %
248: where $\overline N^\cat$ is the total number of events expected in category 
249: $\cat$. 
250: 
251: A total of 68 parameters, including the inclusive signal yield and the
252: parameters from Eq.~(\ref{eq:dt}), are varied in the fit. Most of the
253: parameters describing the continuum distributions are also floated
254: in the fit.
255: 
256: \begin{table*}[t]
257: \begin{center}
258: \input{pdfTable.tex}
259: \vspace{-0.2cm}
260: \caption{ \label{tab:pdfparameterization}
261:         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.}
262: \end{center}
263: \end{table*}
264: 
265: \subsubsection{\boldmath THE $\dt$ AND DALITZ PLOT PDFS}
266: \label{sec:deltaT}
267: 
268: 	The Dalitz plot PDFs require as input the Dalitz plot-dependent 
269: 	relative selection efficiency, $\e=\e(\mprime,\thetaprime)$, 
270: 	and SCF fraction, $\fscf=\fscf(\mprime,\thetaprime)$.
271: 	Both quantities are taken from MC simulation. 
272: 	Away from the Dalitz plot corners the efficiency is uniform, while it 
273: 	decreases when approaching the corners, where one out of the 
274: 	three bodies in the final state is close to rest so that the 
275: 	acceptance requirements on the particle reconstruction become 
276:         restrictive.
277: 	Combinatorial backgrounds and hence SCF fractions are large in
278: 	the corners
279: 	of the Dalitz plot due to the presence of soft neutral clusters 
280: 	and tracks. 
281: 
282: 	For an event~$i$, we define the time-dependent Dalitz plot PDFs
283: 	\beqn
284: 		P_{\tpi-\TM,i} &=&
285: 		\varepsilon_i\,(1 - \fscfi)\,\detJi\,\AmpAll~,
286: 		\\[0.3cm]
287: 		P_{\tpi-\SCF,\,i} &=&
288: 		\varepsilon_i\,\fscfi\,\detJi\,\AmpAll~,
289: 	\eeqn	
290: 	where $P_{\tpi-\TM,i}$ and $P_{\tpi-\SCF,\,i}$ are normalized. The 
291: 	corresponding phase space integration involves the expectation values 	
292: 	$\langle \varepsilon\,(1-\fscf)\,\detJ \,f^\kappa f^{\sigma*}\rangle$
293: 	and 
294: 	$\langle \varepsilon\,\fscf\,\detJ\, f^\kappa f^{\sigma*}\rangle$
295: 	for TM and SCF events, where the indices $\kappa$, $\sigma$ 
296: 	run over all resonances belonging to the signal model.
297: 	The expectation values are model-dependent and are 
298: 	computed with the use of MC integration over the square Dalitz plot:
299: 	\beq
300: 	\label{eq:normAverage}
301: 		\langle \varepsilon\,(1-\fscf)\,\detJ\, f^\kappa f^{\sigma*}\rangle
302: 		\;=\; \frac{\int_0^1\int_0^1 
303: 			    \varepsilon\,(1-\fscf)\,\detJ\, f^\kappa f^{\sigma*}
304: 			\,d\mprime d\thetaprime}
305: 		       {\int_0^1\int_0^1 \varepsilon\,\detJ\, f^\kappa f^{\sigma*}
306: 			\,d\mprime d\thetaprime}~,
307: 	\eeq
308: 	and similarly for 
309: 	$\langle \varepsilon\,\,\detJ\, f^\kappa f^{\sigma*}\rangle$,
310: 	where all quantities in the integrands are Dalitz plot-dependent.
311: 
312: 	Equation~(\ref{eq:theLikelihood}) invokes the phase 
313: 	space-averaged SCF fraction 
314: 	$\fscfave\equiv\langle\fscf\,\detJ\, f^\kappa f^{\sigma*}\rangle$. 
315: 	The PDF normalization  is decay-dynamics-dependent
316: 	and is computed iteratively. We 
317: 	determine the average SCF fractions separately for each tagging category 
318: 	from MC simulation. 
319: 	
320: 	The width of the dominant $\rho(770)$ resonance is large compared 
321: 	to the mass resolution for TM events (about $8\mevcc$ core Gaussian
322: 	resolution). We  therefore neglect resolution effects in the TM 
323: 	model.	
324: 	Misreconstructed events	have a poor mass resolution that strongly 
325: 	varies across the Dalitz plot. It is described in the fit by a 
326: 	$2\times 2$-dimensional resolution function
327: 	\beq
328: 	\label{eq:rscf}
329: 		\Rscf(\mprime_r,\thetaprime_r,\mprime_t,\thetaprime_t)~,
330: 	\eeq
331: 	which represents the probability to reconstruct at the coordinate
332: 	$(\mprime_r,\thetaprime_r)$ an event that has the true coordinate 
333: 	$(\mprime_t,\thetaprime_t)$. It obeys the unitarity condition
334: 	\beq
335: 		\intl_0^1\intl_0^1 
336: 		\Rscf(\mprime_r,\thetaprime_r,\mprime_t,\thetaprime_t)
337: 		\,d\mprime_r d\thetaprime_r = 1,~
338: 	\eeq
339: 	and is convolved with the signal model. 
340: 	The $\Rscf$ function is obtained from MC simulation.
341: 
342: 	We use the signal model described in Section~\ref{sec:kinmeatics}. 
343: 	It contains the dynamical information and is connected with $\dt$ via 
344: 	the matrix element~(\ref{eq:dt}), which serves as PDF. It is diluted 
345: 	by the effects of mistagging and the limited vertex 
346: 	resolution~\cite{rhopipaper}. 
347: 	The $\deltat$ resolution function for signal and \B-background 
348: 	events is a sum of three Gaussian distributions, with parameters 
349: 	determined by a fit to fully reconstructed $\Bz$ 
350: 	decays~\cite{BabarS2b}. 
351: \\[0.3cm]\noindent
352: 	The Dalitz plot- and $\dt$-dependent PDFs factorize for the 
353: 	charged-$B$-background modes, but not (necessarily) 
354: 	for the neutral-$B$ background due to $\BzBzb$ mixing.
355: 
356:  	The charged \B-background
357: 		contribution to the likelihood~(\ref{eq:theLikelihood})
358:                 involves 
359: 		the parameter $\Atag$, multiplied by the tag flavor $\Qtag$ of 
360: 		the event. In the presence of significant tag-``charge'' 
361: 		correlation (represented by an effective 
362: 		flavor-tag-versus-Dalitz-coordinate correlation),
363: 		it parameterizes possible direct \CP violation in these events.
364: 		We also use distinct square Dalitz plot PDFs for each 
365: 		reconstructed $B$ flavor tag, and a flavor-tag-averaged PDF for 
366: 		untagged events. The PDFs are obtained from MC simulation and are 
367: 		described with the use of non-parametric functions.
368: 		The $\dt$ resolution parameters are determined by a fit to fully 
369: 		reconstructed $\Bp$ decays. For each $\Bp$-background class we adjust 
370: 		effective lifetimes to account for the misreconstruction of the 
371: 		event that modifies the nominal $\dt$ resolution function.
372: 
373: 	The neutral-$B$ background is parameterized with PDFs that
374: 		depend on the flavor tag of the event. In the case of \CP
375: 		eigenstates, correlations between the flavor tag and the Dalitz 
376: 		coordinate are expected to be small. However, non-\CP  eigenstates,
377: 		such as $a_1^\pm\pi^\mp$, may exhibit such correlation. Both types 
378: 		of decays can have direct
379: 		and mixing-induced \CP  violation. A third type of decays
380: 		involves charged kaons and does not exhibit mixing-induced
381: 		\CP  violation, but usually has a strong correlation between the
382: 		flavor tag and the Dalitz plot coordinate (the kaon charge), because 
383: 		it consists of $B$-flavor eigenstates.
384: 		The Dalitz plot PDFs are obtained from MC simulation and are 
385: 		described with the use of non-parametric functions.
386: 		For neutral $B$ background, the signal $\dt$ resolution model 
387: 		is assumed.
388: 
389: 	The Dalitz plot
390: 		treatment of the continuum events is similar to the one used
391: 		for charged-$B$ background. 
392: 		The square Dalitz plot PDF for continuum background is 
393: 		obtained from on-resonance events selected in the
394: 		$\mes$ sidebands and corrected for feed-through
395: 		from \B decays. A large number of cross checks has been 
396: 		performed to ensure the high fidelity of the empirical shape 
397: 		parameterization. Analytical models have been found insufficient.
398: 		The continuum $\deltat$ distribution is parameterized as the sum of 
399: 		three Gaussian distributions with common mean and 
400: 		three distinct widths that scale the $\dt$ per-event error. 
401: 		This yields six shape parameters that are determined by 
402: 		the fit.
403:  		The model is motivated by the observation that 
404: 		the $\dt$ average is independent of its error, and that the 
405: 		$\dt$ RMS depends linearly on the $\dt$ error.
406: 
407: \subsubsection{PARAMETERIZATION OF THE OTHER VARIABLES}
408: \label{sec:likmESanddE}
409: 
410: 	The $\mes$ distribution of TM signal events is
411: 		parameterized by a bifurcated Crystal Ball function~\cite{PDFsCB},
412: 		which is a combination of a one-sided Gaussian and 
413: 		a Crystal Ball function. The mean of this function
414: 		is determined by the fit. A non-parametric
415: 		function is used to describe the SCF signal component.
416: 
417: 	The $\deprime$ distribution of TM events is
418: 		parameterized by a double Gaussian function, where
419: 		all five parameters depend linearly on $\mpm^2$.
420: 		Misreconstructed events are parameterized by a broad
421: 		single Gaussian function.
422: 		
423: 	Both $\mes$ and $\deprime$ PDFs are parameterized by non-parametric
424: 		functions for all $B$-background classes.
425: 
426: 	The $\mes$ and $\deprime$ PDFs for continuum events are
427: 		parameterized with an Argus shape function~\cite{PDFsArgus} and 
428: 		a second order polynomial, respectively, with parameters 
429: 		determined by the fit.
430: 
431: 	We use non-parametric functions to empirically describe the distributions 
432: 		of the NN outputs
433: 		found in the MC simulation for TM and SCF signal events, 
434: 		and for \B-background events. We distinguish tagging categories 
435: 		for TM signal events to account for differences observed in the 
436: 		shapes.
437: 	
438: 	The continuum NN distribution is parameterized by a 
439: 		third order polynomial that is defined to be positive. 
440: 		The coefficients of the polynomial are determined by the fit.
441: 		Continuum events exhibit a correlation between the Dalitz plot 
442: 		coordinate
443: 		and the shape of the event that is exploited in the NN. 
444: To correct for residual effects,
445: 		we introduce a linear dependence of the polynomial coefficients
446: 		on the distance of the Dalitz plot coordinate to the kinematic 
447: 		boundaries of the Dalitz plot. The parameters describing this
448: 		dependence are determined by the fit.
449: 
450: