ViewVC Help
View File | Revision Log | Show Annotations | Root Listing
root/cvsroot/UserCode/benhoob/cmsnotes/StopSearch/CR4.tex
Revision: 1.13
Committed: Fri Oct 12 20:09:46 2012 UTC (12 years, 7 months ago) by vimartin
Content type: application/x-tex
Branch: MAIN
Changes since 1.12: +2 -2 lines
Log Message:
typo fixes

File Contents

# User Rev Content
1 claudioc 1.1 \subsection{Dilepton studies in CR4}
2     \label{sec:cr4}
3    
4     \subsubsection{Modeling of Additional Hard Jets in Top Dilepton Events}
5     \label{sec:jetmultiplicity}
6    
7     Dilepton \ttbar\ events have 2 jets from the top decays, so additional
8     jets from radiation or higher order contributions are required to
9 claudioc 1.10 enter the signal sample. In this Section we develop an algorithm
10 vimartin 1.13 to be applied to all \ttll\ MC samples to ensure that the distribution
11 claudioc 1.10 of extra jets is properly modelled.
12    
13    
14     The modeling of additional jets in \ttbar\
15 claudioc 1.1 events is checked in a \ttll\ control sample,
16     selected by requiring
17     \begin{itemize}
18     \item exactly 2 selected electrons or muons with \pt $>$ 20 GeV
19 linacre 1.8 \item \met\ $>$ 50 GeV
20 claudioc 1.1 \item $\geq1$ b-tagged jet
21 burkett 1.2 \item Z-veto ($|m_{\ell\ell} - 91| > 15$ GeV)
22 claudioc 1.1 \end{itemize}
23     Figure~\ref{fig:dileptonnjets} shows a comparison of the jet
24     multiplicity distribution in data and MC for this two-lepton control
25     sample. After requiring at least 1 b-tagged jet, most of the
26     events have 2 jets, as expected from the dominant process \ttll. There is also a
27     significant fraction of events with additional jets.
28     The 3-jet sample is mainly comprised of \ttbar\ events with 1 additional
29     emission and similarly the $\ge4$-jet sample contains primarily
30     $\ttbar+\ge2$ jet events.
31     %Even though the primary \ttbar\
32     %Madgraph sample used includes up to 3 additional partons at the Matrix
33     %Element level, which are intended to describe additional hard jets,
34     %Figure~\ref{fig:dileptonnjets} shows a slight mis-modeling of the
35     %additional jets.
36    
37    
38     \begin{figure}[hbt]
39     \begin{center}
40 linacre 1.8 \includegraphics[width=0.5\linewidth]{plots/njets_all_met50_mueg.pdf}
41     \includegraphics[width=0.5\linewidth]{plots/njets_all_met50_diel.pdf}%
42     \includegraphics[width=0.5\linewidth]{plots/njets_all_met50_dimu.pdf}
43 claudioc 1.1 \caption{
44     \label{fig:dileptonnjets}%\protect
45     Comparison of the jet multiplicity distribution in data and MC for dilepton events in the \E-\M\
46     (top), \E-\E\ (bottom left) and \M-\M\ (bottom right) channels.}
47     \end{center}
48     \end{figure}
49    
50     It should be noted that in the case of \ttll\ events
51     with a single reconstructed lepton, the other lepton may be
52     mis-reconstructed as a jet. For example, a hadronic tau may be
53     mis-identified as a jet (since no $\tau$ identification is used).
54     In this case only 1 additional jet from radiation may suffice for
55     a \ttll\ event to enter the signal sample. As a result, both the
56     samples with $\ttbar+1$ jet and $\ttbar+\ge2$ jets are relevant for
57 burkett 1.2 estimating the top dilepton background in the signal region.
58 claudioc 1.1
59     %In this section we discuss a correction to $ N_{2 lep}^{MC} $ in Equation XXX
60     %due to differences in the modelling of the jet multiplicity in data versus MC.
61     %The same correction also enters $ N_{peak}^{MC}$ in Equation XXX to the extend that the
62     %dilepton contributions to $ N_{peak}^{MC}$ gets corrected.
63    
64     %The dilepton control sample is defined by the following requirements:
65     %\begin{itemize}
66     %\item Exactly 2 selected electrons or muons with \pt $>$ 20 GeV
67     %\item \met\ $>$ 50 GeV
68     %\item $\geq1$ b-tagged jet
69     %\end{itemize}
70     %
71     %This sample is dominated by \ttll. The distribution of \njets\ for data and MC passing this selection is displayed in Fig.~\ref{fig:dilepton_njets}.
72     %We use this distribution to derive scale factors which reweight the \ttll\ MC \njets\ distribution to match the data. We define the following
73     %quantities
74     %
75     %\begin{itemize}
76     %\item $N_{2}=$ data yield minus non-dilepton \ttbar\ MC yield for \njets\ $\leq$ 2
77     %\item $N_{3}=$ data yield minus non-dilepton \ttbar\ MC yield for \njets\ = 3
78     %\item $N_{4}=$ data yield minus non-dilepton \ttbar\ MC yield for \njets\ $\geq$ 4
79     %\item $M_{2}=$ dilepton \ttbar\ MC yield for \njets\ $\leq$ 2
80     %\item $M_{3}=$ dilepton \ttbar\ MC yield for \njets\ = 3
81     %\item $M_{4}=$ dilepton \ttbar\ MC yield for \njets\ $\geq$ 4
82     %\end{itemize}
83     %
84     %We use these yields to define 3 scale factors, which quantify the data/MC ratio in the 3 \njets\ bins:
85     %
86     %\begin{itemize}
87     %\item $SF_2 = N_2 / M_2$
88     %\item $SF_3 = N_3 / M_3$
89     %\item $SF_4 = N_4 / M_4$
90     %\end{itemize}
91     %
92     %And finally, we define the scale factors $K_3$ and $K_4$:
93     %
94     %\begin{itemize}
95     %\item $K_3 = SF_3 / SF_2$
96     %\item $K_4 = SF_4 / SF_2$
97     %\end{itemize}
98     %
99     %The scale factor $K_3$ is extracted from dilepton \ttbar\ events with \njets = 3, which have exactly 1 ISR jet.
100     %The scale factor $K_4$ is extracted from dilepton \ttbar\ events with \njets $\geq$ 4, which have at least 2 ISR jets.
101     %Both of these scale factors are needed since dilepton \ttbar\ events which fall in our signal region (including
102     %the \njets $\geq$ 4 requirement) may require exactly 1 ISR jet, in the case that the second lepton is reconstructed
103     %as a jet, or at least 2 ISR jets, in the case that the second lepton is not reconstructed as a jet. These scale
104     %factors are applied to the dilepton \ttbar\ MC only. For a given MC event, we determine whether to use $K_3$ or $K_4$
105     %by counting the number of reconstructed jets in the event ($N_{\rm{jets}}^R$) , and subtracting off any reconstructed
106     %jet which is matched to the second lepton at generator level ($N_{\rm{jets}}^\ell$); $N_{\rm{jets}}^{\rm{cor}} = N_{\rm{jets}}^R - N_{\rm{jets}}^\ell$.
107     %For events with $N_{\rm{jets}}^{\rm{cor}}=3$ the factor $K_3$ is applied, while for events with $N_{\rm{jets}}^{\rm{cor}}\geq4$ the factor $K_4$ is applied.
108     %For all subsequent steps, the scale factors $K_3$ and $K_4$ have been
109     %applied to the \ttll\ MC.
110    
111    
112     Table~\ref{tab:njetskfactors} shows scale factors ($K_3$ and $K_4$)
113     used to correct the
114     fraction of events with additional jets in MC to the observed fraction
115     in data. These scale factors are calculated from Fig.~\ref{fig:dileptonnjets}
116     as follows:
117     \begin{itemize}
118 claudioc 1.10 \item $N_{2}=$ data yield minus non-dilepton \ttbar\ MC yield for
119     \njets\ =1 or 2.
120 claudioc 1.1 \item $N_{3}=$ data yield minus non-dilepton \ttbar\ MC yield for \njets\ = 3
121     \item $N_{4}=$ data yield minus non-dilepton \ttbar\ MC yield for \njets\ $\geq$ 4
122 claudioc 1.10 \item $M_{2}=$ dilepton \ttbar\ MC yield for \njets\ = 1 or 2
123 claudioc 1.1 \item $M_{3}=$ dilepton \ttbar\ MC yield for \njets\ = 3
124     \item $M_{4}=$ dilepton \ttbar\ MC yield for \njets\ $\geq$ 4
125     \end{itemize}
126     \noindent then
127     \begin{itemize}
128     \item $SF_2 = N_2 / M_2$
129     \item $SF_3 = N_3 / M_3$
130     \item $SF_4 = N_4 / M_4$
131     \item $K_3 = SF_3 / SF_2$
132     \item $K_4 = SF_4 / SF_2$
133     \end{itemize}
134     \noindent This insures that $K_3 M_3/(M_2 + K_3 M_3 + K_4 M_4) = N_3 /
135     (N_2+N_3+N_4)$ and similarly for the $\geq 4$ jet bin.
136    
137 linacre 1.8 Table~\ref{tab:njetskfactors} also shows the values of $K_3$ and $K_4$ when the \met\ cut in the control sample definition is changed from 50 GeV to 100 GeV and 150 GeV.
138 claudioc 1.10 % These values of $K_3$ and $K_4$ are not used in the analysis, but
139 vimartin 1.13 This demonstrates that there is no statistically significant dependence of $K_3$ and $K_4$ on the \met\ cut.
140 claudioc 1.1
141 linacre 1.8
142 claudioc 1.10 The factors $K_3$ and $K_4$ (derived with the 100 GeV \met\ cut) are applied to the \ttll\ MC throughout the
143 claudioc 1.1 entire analysis, i.e.
144     whenever \ttll\ MC is used to estimate or subtract
145 claudioc 1.10 a yield or distribution. To be explicit, whenever Powheg is used,
146     the Powheg $K_3$ and $K_4$ are used; whenever default MadGraph is
147     used, the MadGraph $K_3$ and $K_4$ are used, etc.
148 claudioc 1.1 %
149     In order to do so, it is first necessary to count the number of
150     additional jets from radiation and exclude leptons mis-identified as
151     jets. A jet is considered a mis-identified lepton if it is matched to a
152     generator-level second lepton with sufficient energy to satisfy the jet
153     \pt\ requirement ($\pt>30~\GeV$). Then \ttll\ events that need two
154     radiation jets to enter our selection are scaled by $K_4$,
155     while those that only need one radiation jet are scaled by $K_3$.
156    
157     \begin{table}[!ht]
158     \begin{center}
159 vimartin 1.12 {\footnotesize
160 linacre 1.8 \begin{tabular}{l|c|c|c}
161     \cline{2-4}
162 linacre 1.9 & \multicolumn{3}{c}{ \met\ cut for data/MC scale factors} \\
163 claudioc 1.1 \hline
164 linacre 1.8 Jet Multiplicity Sample & 50 GeV & 100 GeV & 150 GeV \\
165 claudioc 1.1 \hline
166     \hline
167 linacre 1.8 N jets $= 3$ (sensitive to $\ttbar+1$ extra jet from radiation)
168     & $K_3 = 0.98 \pm 0.02$ & $K_3 = 1.01 \pm 0.03$ & $K_3 = 1.00 \pm 0.08$ \\
169 claudioc 1.1 N jets $\ge4$ (sensitive to $\ttbar+\ge2$ extra jets from radiation)
170 linacre 1.8 & $K_4 = 0.94 \pm 0.02$ & $K_4 = 0.93 \pm 0.04$ & $K_4 = 1.00 \pm 0.08$ \\
171 claudioc 1.1 \hline
172 vimartin 1.12 \end{tabular}}
173 claudioc 1.1 \caption{Data/MC scale factors used to account for differences in the
174     fraction of events with additional hard jets from radiation in
175 claudioc 1.10 \ttll\ events. The values derived with the 100 GeV \met\ cut are applied
176 linacre 1.8 to the \ttll\ MC throughout the analysis. \label{tab:njetskfactors}}
177 claudioc 1.1 \end{center}
178     \end{table}
179    
180     \clearpage
181    
182    
183    
184     \subsubsection{Validation of the ``Physics'' Modelling of the \ttdl\
185     MC in CR4}
186 burkett 1.2 \label{sec:CR4-valid}
187 claudioc 1.1
188     As mentioned above, $t\bar{t} \to $ dileptons where one of the leptons
189     is somehow lost constitutes the main background.
190     The object of this test is to validate the $M_T$ distribution of this
191     background by looking at the $M_T$ distribution of well identified
192     dilepton events.
193     We construct a transverse mass variable from the leading lepton and
194 vimartin 1.4 the \met. We distinguish between events with leading electrons and
195 claudioc 1.1 leading muons.
196    
197     The $t\bar{t}$ MC is corrected using the $K_3$ and $K_4$ factors
198     from Section~\ref{sec:jetmultiplicity}. It is also normalized to the
199 claudioc 1.11 total data yield separately for the \met\ requirements of the various signal
200     regions. These normalization factors are listed
201 claudioc 1.1 in Table~\ref{tab:cr4mtsf} and are close to unity.
202    
203     The underlying \met\ and $M_T$ distributions are shown in
204 burkett 1.2 Figures~\ref{fig:cr4met} and~\ref{fig:cr4mtrest}. The data-MC agreement
205 claudioc 1.1 is quite good. Quantitatively, this is also shown in Table~\ref{tab:cr4yields}.
206 claudioc 1.10 This is a {\bf very} important Table. It shows that for well
207     identified \ttdl\ , the MC can predict the $M_T$ tail. Since the
208     main background is also \ttdl\ except with one ``missed'' lepton,
209     this is a key test.
210 claudioc 1.1
211     \begin{table}[!h]
212     \begin{center}
213 vimartin 1.3 {\footnotesize
214 vimartin 1.6 \begin{tabular}{l||c||c|c|c|c|c|c}
215 claudioc 1.1 \hline
216 vimartin 1.3 Sample & CR4PRESEL & CR4A & CR4B & CR4C &
217 vimartin 1.6 CR4D & CR4E & CR4F\\
218 claudioc 1.1 \hline
219     \hline
220 vimartin 1.6 $\mu$ Data/MC-SF & $1.01 \pm 0.03$ & $0.96 \pm 0.04$ & $0.99 \pm 0.07$ & $1.05 \pm 0.13$ & $0.91 \pm 0.20$ & $1.10 \pm 0.34$ & $1.50 \pm 0.67$ \\
221 claudioc 1.1 \hline
222     \hline
223 vimartin 1.6 e Data/MC-SF & $0.99 \pm 0.03$ & $0.99 \pm 0.05$ & $0.91 \pm 0.08$ & $0.84 \pm 0.13$ & $0.70 \pm 0.18$ & $0.73 \pm 0.29$ & $0.63 \pm 0.38$ \\
224 claudioc 1.1 \hline
225 vimartin 1.3 \end{tabular}}
226 claudioc 1.1 \caption{ Data/MC scale factors for total yields, applied to compare
227     the shapes of the distributions.
228     The uncertainties are statistical only.
229     \label{tab:cr4mtsf}}
230     \end{center}
231     \end{table}
232    
233    
234     \begin{table}[!h]
235     \begin{center}
236 vimartin 1.3 {\footnotesize
237 vimartin 1.6 \begin{tabular}{l||c||c|c|c|c|c|c}
238 claudioc 1.1 \hline
239 vimartin 1.3 Sample & CR4PRESEL & CR4A & CR4B & CR4C &
240 vimartin 1.6 CR4D & CR4E & CR4F\\
241 claudioc 1.1 \hline
242     \hline
243 vimartin 1.7 $\mu$ MC & $256 \pm 14$ & $152 \pm 11$ & $91 \pm 9$ & $26 \pm 5$ & $6 \pm 2$ & $4 \pm 2$ & $2 \pm 1$ \\
244 vimartin 1.6 $\mu$ Data & $251$ & $156$ & $98$ & $27$ & $8$ & $6$ & $4$ \\
245 claudioc 1.1 \hline
246 vimartin 1.7 $\mu$ Data/MC SF & $0.98 \pm 0.08$ & $1.02 \pm 0.11$ & $1.08 \pm 0.16$ & $1.04 \pm 0.28$ & $1.29 \pm 0.65$ & $1.35 \pm 0.80$ & $2.10 \pm 1.72$ \\
247 claudioc 1.1 \hline
248     \hline
249 vimartin 1.7 e MC & $227 \pm 13$ & $139 \pm 11$ & $73 \pm 8$ & $21 \pm 4$ & $5 \pm 2$ & $2 \pm 1$ & $1 \pm 1$ \\
250 vimartin 1.6 e Data & $219$ & $136$ & $72$ & $19$ & $2$ & $1$ & $1$ \\
251 claudioc 1.1 \hline
252 vimartin 1.7 e Data/MC SF & $0.96 \pm 0.09$ & $0.98 \pm 0.11$ & $0.99 \pm 0.16$ & $0.92 \pm 0.29$ & $0.41 \pm 0.33$ & $0.53 \pm 0.62$ & $0.76 \pm 0.96$ \\
253     \hline
254     \hline
255     $\mu$+e MC & $483 \pm 19$ & $291 \pm 16$ & $164 \pm 13$ & $47 \pm 7$ & $11 \pm 3$ & $6 \pm 2$ & $3 \pm 2$ \\
256     $\mu$+e Data & $470$ & $292$ & $170$ & $46$ & $10$ & $7$ & $5$ \\
257     \hline
258     $\mu$+e Data/MC SF & $0.97 \pm 0.06$ & $1.00 \pm 0.08$ & $1.04 \pm 0.11$ & $0.99 \pm 0.20$ & $0.90 \pm 0.37$ & $1.11 \pm 0.57$ & $1.55 \pm 1.04$ \\
259 claudioc 1.1 \hline
260 vimartin 1.3 \end{tabular}}
261 claudioc 1.1 \caption{ Yields in \mt\ tail comparing the MC prediction (after
262     applying SFs) to data. The uncertainties are statistical only.
263     \label{tab:cr4yields}}
264     \end{center}
265     \end{table}
266    
267     \begin{figure}[hbt]
268     \begin{center}
269     \includegraphics[width=0.5\linewidth]{plots/CR4plots/met_met50_leadmuo_nj4.pdf}%
270     \includegraphics[width=0.5\linewidth]{plots/CR4plots/met_met50_leadele_nj4.pdf}
271     \includegraphics[width=0.5\linewidth]{plots/CR4plots/mt_met100_leadmuo_nj4.pdf}%
272     \includegraphics[width=0.5\linewidth]{plots/CR4plots/mt_met100_leadele_nj4.pdf}
273     \caption{
274     Comparison of the \met\ (top) and \mt\ for $\met>100$ (bottom) distributions in data vs. MC for events
275     with a leading muon (left) and leading electron (right)
276     satisfying the requirements of CR4.
277     \label{fig:cr4met}
278     }
279     \end{center}
280     \end{figure}
281    
282     \begin{figure}[hbt]
283     \begin{center}
284 vimartin 1.3 \includegraphics[width=0.5\linewidth]{plots/CR4plots/mt_met50_leadmuo_nj4.pdf}%
285     \includegraphics[width=0.5\linewidth]{plots/CR4plots/mt_met50_leadele_nj4.pdf}
286 claudioc 1.1 \includegraphics[width=0.5\linewidth]{plots/CR4plots/mt_met150_leadmuo_nj4.pdf}%
287     \includegraphics[width=0.5\linewidth]{plots/CR4plots/mt_met150_leadele_nj4.pdf}
288     \includegraphics[width=0.5\linewidth]{plots/CR4plots/mt_met200_leadmuo_nj4.pdf}%
289     \includegraphics[width=0.5\linewidth]{plots/CR4plots/mt_met200_leadele_nj4.pdf}
290     \caption{
291     Comparison of the \mt\ distribution in data vs. MC for events
292     with a leading muon (left) and leading electron (right)
293     satisfying the requirements of CR4. The \met\ requirements used are
294 vimartin 1.3 50 GeV (top), 200 GeV (middle) and 250 GeV (bottom).
295 claudioc 1.1 \label{fig:cr4mtrest}
296     }
297     \end{center}
298     \end{figure}
299    
300    
301     \clearpage