claudioc/OSNote2010/otherBG.tex

\section{Non $t\bar{t}$ Backgrounds}
\label{sec:othBG}

Backgrounds from divector bosons and single top
can be reliably estimated from Monte Carlo.
They are negligible compared to $t\bar{t}$. 
 
%Backgrounds from Drell Yan are also expected
%to be negligible from MC.  However one always 
%worries about the modeling of tails of the \met. 
%In the context of other dilepton analyses we 
%have developed a data driven method to estimate 
%the number of Drell Yan events\cite{ref:dy}.
%The method is based on counting the number of
%$Z$ candidates passing the full selection, and 
%then scaling by the expected ratio of Drell Yan
%events outside vs. inside the $Z$ mass 
%window.\footnote{A correction based on $e\mu$ events 
%is also applied.}  This ratio is called $R_{out/in}$
%and is obtained from Monte Carlo.

%To estimate the Drell-Yan contribution in the four $ABCD$ 
%regions, we count the numbers of $Z \to ee$ and $Z \to \mu\mu$
%events falling in each region, we subtract off the number 
%of $e\mu$ events with $76 < M(e\mu) < 106$ GeV, and 
%we multiply the result by $R_{out/in}$ from Monte Carlo.
%The results are summarized in Table~\ref{tab:ABCD-DY}.

%\begin{table}[hbt]
%\begin{center}
%\caption{\label{tab:ABCD-DY} Drell-Yan estimations
%in the four 
%regions of Figure~\ref{fig:abcdData}.  The yields are
%for dileptons with invariant mass consistent with the $Z$.
%The factor
%$R_{out/in}$ is from MC.  All uncertainties 
%are statistical only.  In regions $A$ and $D$ there is no statistics
%in the Monte Carlo to calculate $R_{out/in}$.}
%\begin{tabular}{|l|c|c|c||c|}
%\hline
%Region   & $N(ee)+N(\mu\mu)$ & $N(e\mu)$ & $R_{out/in}$ & Estimated DY BG \\
%\hline
%$A$      &  0                & 0         & ??          & ??$\pm$xx   \\
%$B$      &  5                & 1         & 2.5$\pm$1.0 & 9$\pm$xx   \\
%$C$      &  0                & 0         & 1.$\pm$1.   & 0$\pm$xx   \\
%$D$      &  0                & 0         & ??          & 0$\pm$xx   \\
%\hline
%\end{tabular}
%\end{center}
%\end{table}


%When  find no dilepton events with invariant mass
%consistent with the $Z$ in the signal region.
%Using the value of 0.1 for the ratio described above, this 
%means that the Drell Yan background in our signal 
%region is $< 0.23\%$ events at the 90\% confidence level.
%{\color{red} (If we find 1 event this will need to be adjusted)}.

As discussed in Section~\ref{sec:victory}, residual Drell-Yan 
events can have a significant effect on the data driven background
prediction based on $P_T(\ell\ell)$.  This is taken into account,
based on MC expectations, 
by the $K_C$ factor described in that Section.  
As a cross-check, we use a separate data driven method to 
estimate the impact of Drell Yan events on the 
background prediction based on $P_T(\ell\ell)$. 
In this method\cite{ref:top} we count the number 
of $Z$ candidates\footnote{$e^+e^-$ and $\mu^+\mu^-$
with invariant mass between 76 and 106 GeV.}
passing the same selection as 
region $D$ except that the $\met/\sqrt{\rm SumJetPt}$ requirement is 
replaced by a $P_T(\ell\ell)/\sqrt{\rm SumJetPt}$ requirement
(we call this the ``region $D'$ selection'').
We subtract off the number of $e\mu$ events with 
invariant mass in the $Z$ passing the region $D'$ selection.
Finally, we multiply the result by ratio $R_{out/in}$ derived
from Monte Carlo as the ratio of Drell-Yan events
outside/inside the $Z$ mass window 
in the $D'$ region.

We find $N^{D'}(ee+\mu\mu)=2$, $N^{D'}(e\mu)=0$,
$R^{D'}_{out/in}=0.4\pm0.3$ (stat.). 
Thus we estimate the number of Drell Yan events in region $D'$ to 
be $0.8\pm X$. 
{\color{red} \bf Update Rout/in with dpt/pt cut and Zmumugamma veto}

This Drell Yan method could also be used to estimate 
the number of DY events in the signal region (region $D$).
However, there is not enough statistics in the Monte 
Carlo to make a measurement of $R_{out/in}$ in region 
$D$.  In any case, no $Z \to \ell\ell$ candidates are
found in region $D$.


%In the context of other dilepton analyses we 
%have developed a data driven method to estimate 
%the number of Drell Yan events\cite{ref:dy}.
%The method is based on counting the number of
%$Z$ candidates passing the full selection, and 
%then scaling by the expected ratio of Drell Yan
%events outside vs. inside the $Z$ mass 
%window.\footnote{A correction based on $e\mu$ events 
%is also applied.}  This ratio is called $R_{out/in}$
%and is obtained from Monte Carlo.


%As a cross-check, we use the same Drell Yan background
%estimation method described above to estimate the 
%number of DY events in the regions $A'B'C'D'$.
%The region $A'$ is defined in the same way as the region $A$ 
%except that the $\met/\sqrt{\rm SumJetPt}$ requirement is 
%replaced by a $P_T(\ell\ell)/\sqrt{\rm SumJetPt}$ requirement.
%The regions B', 
%C', and D' are defined in a similar way.  The results are
%summarized in Table~\ref{tab:ABCD-DYptll}.

%\begin{table}[hbt]
%\begin{center}
%\caption{\label{tab:ABCD-DYptll} Drell-Yan estimations
%in the four 
%regions $A'B'C'D'$ defined in the text.  The yields are
%for dileptons with invariant mass consistent with the $Z$.
%The factor
%$R_{out/in}$ is from MC.  All uncertainties 
%are statistical only.}
%\begin{tabular}{|l|c|c|c||c|}
%\hline
%Region    & $N(ee)+N(\mu\mu)$ & $N(e\mu)$ & $R_{out/in}$ & Estimated DY BG \\
%\hline
%$A'$      &  3               & 0        & 0.7$\pm$0.3   & 2.1$\pm$xx   \\
%$B'$      &  3               & 0        & 2.5$\pm$2.1   & 7.5$\pm$xx   \\
%$C'$      &  0               & 0        & 0.1$\pm$0.1   & 0$\pm$xx   \\
%$D'$      &  1               & 0        & 0.4$\pm$0.3   & 0.4$\pm$xx   \\
%\hline
%\end{tabular}
%\end{center}
%\end{table}


Finally, we can use the ``Fake Rate'' method\cite{ref:FR} 
to predict 
the number of events with one fake lepton. We select
events where one of the leptons passes the full selection and 
the other one fails the full selection but passes the 
``Fakeable Object'' selection of 
Reference~\cite{ref:FR}.\footnote{For electrons we use
the V3 fakeable object definition to avoid complications 
associated with electron ID cuts applied in the trigger.} 
We then weight each event passing the full selection
by FR/(1-FR) where FR is the ``fake rate'' for the 
fakeable object.  {\color{red} \bf The results are...} 

{\color{red} \bf We will do the same thing that we did for 
the top analysis, but we will only do it on the full dataset.}
Revision:	1.7
Committed:	Thu Nov 11 15:47:18 2010 UTC (14 years, 5 months ago) by benhoob
Content type:	application/x-tex
Branch:	MAIN
Changes since 1.6:	+5 -5 lines
Log Message:	Minor updates
#	User	Rev	Content
1	claudioc	1.1	\section{Non $t\bar{t}$ Backgrounds}
2			\label{sec:othBG}
3
4			Backgrounds from divector bosons and single top
5			can be reliably estimated from Monte Carlo.
6			They are negligible compared to $t\bar{t}$.
7
8	claudioc	1.4	%Backgrounds from Drell Yan are also expected
9			%to be negligible from MC. However one always
10			%worries about the modeling of tails of the \met.
11			%In the context of other dilepton analyses we
12			%have developed a data driven method to estimate
13			%the number of Drell Yan events\cite{ref:dy}.
14			%The method is based on counting the number of
15			%$Z$ candidates passing the full selection, and
16			%then scaling by the expected ratio of Drell Yan
17			%events outside vs. inside the $Z$ mass
18			%window.\footnote{A correction based on $e\mu$ events
19			%is also applied.} This ratio is called $R_{out/in}$
20			%and is obtained from Monte Carlo.
21
22			%To estimate the Drell-Yan contribution in the four $ABCD$
23			%regions, we count the numbers of $Z \to ee$ and $Z \to \mu\mu$
24			%events falling in each region, we subtract off the number
25			%of $e\mu$ events with $76 < M(e\mu) < 106$ GeV, and
26			%we multiply the result by $R_{out/in}$ from Monte Carlo.
27			%The results are summarized in Table~\ref{tab:ABCD-DY}.
28
29			%\begin{table}[hbt]
30			%\begin{center}
31			%\caption{\label{tab:ABCD-DY} Drell-Yan estimations
32			%in the four
33			%regions of Figure~\ref{fig:abcdData}. The yields are
34			%for dileptons with invariant mass consistent with the $Z$.
35			%The factor
36			%$R_{out/in}$ is from MC. All uncertainties
37			%are statistical only. In regions $A$ and $D$ there is no statistics
38			%in the Monte Carlo to calculate $R_{out/in}$.}
39			%\begin{tabular}{\|l\|c\|c\|c\|\|c\|}
40			%\hline
41			%Region & $N(ee)+N(\mu\mu)$ & $N(e\mu)$ & $R_{out/in}$ & Estimated DY BG \\
42			%\hline
43			%$A$ & 0 & 0 & ?? & ??$\pm$xx \\
44			%$B$ & 5 & 1 & 2.5$\pm$1.0 & 9$\pm$xx \\
45			%$C$ & 0 & 0 & 1.$\pm$1. & 0$\pm$xx \\
46			%$D$ & 0 & 0 & ?? & 0$\pm$xx \\
47			%\hline
48			%\end{tabular}
49			%\end{center}
50			%\end{table}
51	claudioc	1.2
52
53
54			%When find no dilepton events with invariant mass
55			%consistent with the $Z$ in the signal region.
56			%Using the value of 0.1 for the ratio described above, this
57			%means that the Drell Yan background in our signal
58			%region is $< 0.23\%$ events at the 90\% confidence level.
59			%{\color{red} (If we find 1 event this will need to be adjusted)}.
60
61			As discussed in Section~\ref{sec:victory}, residual Drell-Yan
62			events can have a significant effect on the data driven background
63			prediction based on $P_T(\ell\ell)$. This is taken into account,
64			based on MC expectations,
65	benhoob	1.5	by the $K_C$ factor described in that Section.
66	claudioc	1.4	As a cross-check, we use a separate data driven method to
67			estimate the impact of Drell Yan events on the
68			background prediction based on $P_T(\ell\ell)$.
69			In this method\cite{ref:top} we count the number
70			of $Z$ candidates\footnote{$e^+e^-$ and $\mu^+\mu^-$
71			with invariant mass between 76 and 106 GeV.}
72			passing the same selection as
73			region $D$ except that the $\met/\sqrt{\rm SumJetPt}$ requirement is
74			replaced by a $P_T(\ell\ell)/\sqrt{\rm SumJetPt}$ requirement
75			(we call this the ``region $D'$ selection'').
76			We subtract off the number of $e\mu$ events with
77			invariant mass in the $Z$ passing the region $D'$ selection.
78			Finally, we multiply the result by ratio $R_{out/in}$ derived
79			from Monte Carlo as the ratio of Drell-Yan events
80			outside/inside the $Z$ mass window
81			in the $D'$ region.
82
83	benhoob	1.6	We find $N^{D'}(ee+\mu\mu)=2$, $N^{D'}(e\mu)=0$,
84			$R^{D'}_{out/in}=0.4\pm0.3$ (stat.).
85			Thus we estimate the number of Drell Yan events in region $D'$ to
86	benhoob	1.7	be $0.8\pm X$.
87			{\color{red} \bf Update Rout/in with dpt/pt cut and Zmumugamma veto}
88	claudioc	1.4
89			This Drell Yan method could also be used to estimate
90			the number of DY events in the signal region (region $D$).
91			However, there is not enough statistics in the Monte
92			Carlo to make a measurement of $R_{out/in}$ in region
93			$D$. In any case, no $Z \to \ell\ell$ candidates are
94			found in region $D$.
95
96
97
98			%In the context of other dilepton analyses we
99			%have developed a data driven method to estimate
100			%the number of Drell Yan events\cite{ref:dy}.
101			%The method is based on counting the number of
102			%$Z$ candidates passing the full selection, and
103			%then scaling by the expected ratio of Drell Yan
104			%events outside vs. inside the $Z$ mass
105			%window.\footnote{A correction based on $e\mu$ events
106			%is also applied.} This ratio is called $R_{out/in}$
107			%and is obtained from Monte Carlo.
108
109
110
111
112			%As a cross-check, we use the same Drell Yan background
113			%estimation method described above to estimate the
114			%number of DY events in the regions $A'B'C'D'$.
115			%The region $A'$ is defined in the same way as the region $A$
116			%except that the $\met/\sqrt{\rm SumJetPt}$ requirement is
117			%replaced by a $P_T(\ell\ell)/\sqrt{\rm SumJetPt}$ requirement.
118			%The regions B',
119			%C', and D' are defined in a similar way. The results are
120			%summarized in Table~\ref{tab:ABCD-DYptll}.
121
122			%\begin{table}[hbt]
123			%\begin{center}
124			%\caption{\label{tab:ABCD-DYptll} Drell-Yan estimations
125			%in the four
126			%regions $A'B'C'D'$ defined in the text. The yields are
127			%for dileptons with invariant mass consistent with the $Z$.
128			%The factor
129			%$R_{out/in}$ is from MC. All uncertainties
130			%are statistical only.}
131			%\begin{tabular}{\|l\|c\|c\|c\|\|c\|}
132			%\hline
133			%Region & $N(ee)+N(\mu\mu)$ & $N(e\mu)$ & $R_{out/in}$ & Estimated DY BG \\
134			%\hline
135			%$A'$ & 3 & 0 & 0.7$\pm$0.3 & 2.1$\pm$xx \\
136			%$B'$ & 3 & 0 & 2.5$\pm$2.1 & 7.5$\pm$xx \\
137			%$C'$ & 0 & 0 & 0.1$\pm$0.1 & 0$\pm$xx \\
138			%$D'$ & 1 & 0 & 0.4$\pm$0.3 & 0.4$\pm$xx \\
139			%\hline
140			%\end{tabular}
141			%\end{center}
142			%\end{table}
143	claudioc	1.1
144
145			Finally, we can use the ``Fake Rate'' method\cite{ref:FR}
146			to predict
147			the number of events with one fake lepton. We select
148			events where one of the leptons passes the full selection and
149			the other one fails the full selection but passes the
150			``Fakeable Object'' selection of
151			Reference~\cite{ref:FR}.\footnote{For electrons we use
152			the V3 fakeable object definition to avoid complications
153			associated with electron ID cuts applied in the trigger.}
154			We then weight each event passing the full selection
155			by FR/(1-FR) where FR is the ``fake rate'' for the
156	benhoob	1.7	fakeable object. {\color{red} \bf The results are...}
157	claudioc	1.4
158	benhoob	1.7	{\color{red} \bf We will do the same thing that we did for
159			the top analysis, but we will only do it on the full dataset.}