Objectives

• Perform a check to see if data in the dilepton-dijet channel is consistent with SM predictions.
  • This means all we care to do is a hypothesis test that rules out standard model, if we are lucky.
  • Fitting a mass, or measuring a cross section are secondary concerns that should not drive the decision as to what statistical test to use.

Strategy

Traits of non-SM Physics

• Larger than expected number of events with dilepton mass in our target region
• Large dilepton mass values, the larger the less likely it is to be standard model.
• Events clumped together in dilepton mass, indicating a resonance.
• Dijet mass signature reminiscent of a heavy boson.

Basic Strategy

• Model Dilepton and Dijet Mass distributions for signal and background from MC.
• Scan over (many) possible central values for Dilepton Mass (i.e., many different possible Z' masses)
  • How many? Probably no more than one every 10, or even 20GeV. Basically, a binning of one sigma in resolution is probably good enough.
• Create a likelihood ratio using both the Dilepton Mass and Dijet Mass distributions.
• Compare that likelihood to background to produce a probability that the data is from background.
• Run multiple toy MCs with different signal levels to provide a cross-section limit at 95%

LR Strategy

• Define our LR as P(signal)/P(background)
  • Should show marked difference in areas with extremely low background probabilities. Sensitivity to high dilepton masses.
  • Should have little to no effect over Dijet Mass.

We propose this rather than P(signal)/(P(signal)+P(bkg)) because bkg is an exponentially falling shape. Signal events in the half of the gaussian that has larger Mll are thus much less likely to come from background than the ones in the lower half. Ps/Pbkg exploits that better than Ps/(Ps+Pbkg).

Questions

• Do we have to demonstrate how shape changes for the reconstruction Dilepton Mass distribution between 400 and 700 GeV? ?
• How many categories should we split the data up into?
  • Frank's already pointed out that the data from TCE-TCE should look different from some other channels. How much should we do with that?
    • Based on the Mll plots for TCE-TCE, CMUP-CMUP, and CMX-CMX, I suggest to make just two distinctions: electrons vs muons.
      • on second thought, let's look at TCE-Crktrk to figure out if this falls into muon of electron category.
    • In contrast to lllnu analysis, bkg are dominated by real off-shell Z. The only thing that is relevant between different types of ll combos is the Mll resolution.
• How do we do systematics for high Z-masses?
  • Check the CDF Z' search in the dilepton channel to see what they did. Your bkg is the same as theirs, so doing the same, or similar is appropriate.

Basic Plots

Most of these are for me, since they are all in .eps format, and those of you with inferior operating systems will have difficulty reading them. However, I need them in .eps format eventually, so that's more or less what happened.

Run 1.0.0

• Arbitrarily defined control region as having a dilepton mass ob 106 - 400 GeV?
• Arbitrarily defined signal region as having a dilepton mass above 400 GeV?

Basic Plots

• Signal Region
  • Dijet Mass Plot
  • Dilepton Pt Plot
  • Dilepton Mass Plot

• Signal Region
  • Dijet Mass Plot
  • Dilepton Pt Plot
  • Dilepton Mass Plot

Lepton Category plots

I created a basic template for the signal based on the entirety of the WZPrime sample in the signal region. I then compare the function from that fit with three different lepton categories: TCE-TCE, CMUP-CMUP, and CMX-CMX

• Total Fit
• Fit vs. TCE only
  • You should fit the this using the Xtal ball lineshape. It's a shape that includes the radiative tail for electrons. Not sure if this is a default option in root.
• Fit vs. CMUP only
• Fit vs. CMX only
  • Fit the muon distributions with a gaussian. Can't see from the plots if they both have the same resolution. Though they look quite similar.

Background Plots

Early attempts at a background fit follow. It's not very good, and I'll have to work on improving it. It gives virtually no background in the actual signal region.

• Back Fit
  • You shouldn't try to fit the full range from 100GeV. I'd fit from 300-400GeV, and then see how well that matches to the above 400GeV bkg distribution.
    • If you don't have enough statistics in the bkg MC, then start at 250GeV with your fit.

To Be Completed

MC

• Possibly another MC point?

LR

• Code needs to be written and debugged.

Notes

These are for me. Please do not delete them

Ntuples

.x ./Diboson/macros/process.C("/data3/mnorman/Stntuples/wewkwl/stnmaker_prod-prod-xewkwl.000*.s", "test", 2, 9999999,1,999999)

-- MatthewNorman - 11 Jul 2008