Validation Regions for non-prompt background estimation in same charged \(W^{\pm}W^{\pm}\) scattering
(Status report)
Sebastian Ordoñez
jsordonezs@unal.edu.co
23rd of March 2021
Background Estimation
Typically for data-driven background estimation models reliying on statistically independent control (CR), validation (VR) and signal (SR) regions are employed.
- CR: enriched in non-prompt background and extrapolated to SR using a fake factor (FF) -> Max Dilepton CR
VR
SR
CR
FF apply
FF test
- SR: phase space that is defined through selections on kinematic variables, enriched in potential signal of interest -> ssWW
- VR: regions in phase space bet CR and SR where extrapolation is verified->My work
- Validation Region for muons
- Validation Region for electrons
Data-driven Matrix Method
We introduce the following four categories for our analysis. In truth level we define:
- Prompt leptons: originated from \(W^{\pm}\) or \(Z\) decay.
- Non-prompt leptons: coming from other sources e.g. hadron decays faking a signal lepton.
In reco level we have:
- Ana leptons: quite likely to be prompt. Tight kinematic and qualitative criteria, signal leptons in SR.
- Non-Ana leptons: kinematically close to Ana leptons but more likely to be non-promt, i.e. looser object quality selection.
Validation Region
-
Statistical significance: good statistics, i.e. large number of total events and high purity in non-promt events.
-
Composition of the non-promt leptons: in order to guarantee the assumption of FF being the same in both regions.
-
Orthogonality to the signal region.
-
Data modeling: Data in VR sufficiently well modeled by MC.
-
The validation region consists of Ana and Non-Ana leptons and it should be dominated by non-promt events.
-
In order to evaluate the validation region, the following criteria must to be taken into account:
Three lepton muon VR
-
As final state we have one electron and two muons
- Electron Ana selected and at least one muon Ana selected.
- We expect the electron to be prompt and one muon to be prompt too.
- Non Prompt Muons in the SR are a motivation to define \(t\bar{t}\) validation regions for muons
Three lepton muon VR
- We are using now the full Run 2.
- The current cutflow is composed by:
| -Exactly three leptons -At least two Ana leptons |
| Two muons One electron |
| Electron is Ana |
| Two muons with the identical electrical charge |
-
Three leptons
-
Lepton type
-
Ana electron
-
Same charged muons
Three lepton muon VR
By applying the cutflow we have found high purity in non-prompt events. However, we note that low statistics could be a problem.
Three lepton muon VR
Ana electron cut
Three lepton muon VR
Lepton type cut
Three lepton muon VR
Three lepton cut
Classification of non-prompt muons in the VR for events with at least two Ana leptons.
Classification of non-prompt muons in the VR for events with at least two Non-Ana leptons.
Most non-prompt muons come from b-hadron decay
Classification of non-prompt muons in the VR for events with at least two Ana leptons.
Classification of non-prompt muons in the VR for events with at least two Non-Ana leptons.
Classification of non-prompt muons in the VR for events with at least two Ana leptons.
Classification of non-prompt muons in the VR for events with at least two Non-Ana leptons.
Classification of non-prompt muons in the VR for events with at least two Ana leptons.
Classification of non-prompt muons in the VR for events with at least two Non-Ana leptons.
Let us take a closer look to a physical distribution including data and Ana events splitted.
MET for Ana and Non-Ana events
MET for only Ana events
Three lepton muon VR
Three lepton muon VR
MET for Ana and Non-Ana events
MET for only Ana events
Ana electron cut
Three lepton muon VR
MET for Ana and Non-Ana events
MET for only Ana events
Lepton type cut
Three lepton muon VR
MET for Ana and Non-Ana events
MET for only Ana events
Three lepton cut
Conclusions
- We have proven that most non-prompt muons come from b-hadron decay, which is in agreement with to our ssWW Signal Region.
- We have found that data modelling is good.
- The low statistics is being a big problem!
Next steps
-
Since low statistics is our main issue, we have to look for alternatives in our cutflow
-
Adjust the same charge cut (Next slide):
- There are some good options to consider, e.g. the opposite sign cut used by Heng and Joany.
- We can employ a "merged cut" between opposite sign and same charged.
-
Adjust the same charge cut (Next slide):
- Study the composition and the data modelling for these new cuts.
- Adjust everything to apply the fake factors.
Next steps
