Here I show data / simulation comparisons for a subset of the reproduced FF data using a new trigger classification scheme ...

 

The past several data / simulation comparisons I have done for the RFF data set have divided jets into three trigger catagories: jets which pass L2JetHigh, jets which pass JP1, and jets which do not pass either requirement. I divided the jets like this because it was an easy way to take into account the fact that the JP1 trigger was prescaled in data but not in simulation. This scheme will double count jets which fire both triggers and so is not how I want to divide things up for my actual analysis.

 

For this data / simu analysis, I want to test a trigger scheme which I may use in my actual analysis. To do this, I am looking at ~100 runs from the newly reproduced FF data set and the full FF embedding sample. For this test, I am only looking at the CDF Midpoint Cone Algorithm with Radius=0.7 and > 12 TPC hit points.

 

I divide jets into six different catagories defined as:

 

L2JetHigh: The L0 L2JetHigh trigger must have fired (trigger id 240650, 240651, or 240652) and the shouldFire() and didFire() conditions must be met (for simu, only the shouldFire() is needed). Next, the event must satisfy the random, monojet, or dijet conditions of the L2Jet filter. Next, the jet must point to a jet patch above the JP-th2 threshold or to two adjacent jet patches both of which are above the JP-th0 threshold. Finally, the jet must have a transverse momentum greater than 8.4.

 

JP1Lo: The L0 JP1 trigger must have fired (trigger id 240410 or 240411) and the shouldFire() and didFire() conditions must be met (for simu, only the shouldFire() is needed). Next, the jet must point to a jet patch above the JP-th1 threshold. Finally, the jet must have a transverse momentum less than 8.4.

 

JP1Hi: This catagory is the same as JP1Lo except for the following two points: The jet must have transverse momentum greater than 8.4 and the condition '!(jp2 && jp2->didFire() && jp2->shouldFire())' must be satisfied. Here jp2 is the L0 L2JetHigh trigger.

 

UnL2JetHigh: These jets do not meet the requirements for L2JetHigh, JP1Lo, or JP1Hi but the other jet in the dijet event met the requirements for L2JetHigh.

 

UnJP1Lo: These jets do not meet the requirements for L2JetHigh, JP1Lo, or JP1Hi but the other jet in the dijet event met the requirements for JP1Lo.

 

UnJP1Hi: These jets do not meet the requirements for L2JetHigh, JP1Lo, or JP1Hi but the other jet in the dijet event met the requirements for JP1Hi.

 

I can now compare data to simulation for single jet quantities such as pt, eta, phi, etc. for each trigger.

 

Figure 1: The top six pannels of this figure show the pt spectra of the high pt jets in the dijet. The blue curves show data and the red curves show simulation. From left to right, top to bottom, the pannels are: L2JetHigh, JP1Lo, JP1Hi, UnL2JetHigh, UnJP1Lo, and UnJP1Hi. The bottom six pannels show the corresponding data / simulation ratios.

 

Figure 2: This figure has the same layout as figure 1, but now I show the eta spectra of the high pt jets in the dijet.

 

When looking at dijet quantities, such as dijet invariant mass, which include both jets, we need to look at all combinations of the above 6 triggers. In general, there are 36 trigger combinations (2 jets and 6 triggers) but by definition, a valid event must have at least one triggered jet, so there can be no untriggered-untriggered events, this eliminates 9 combinations. Furthermore, the untriggered jets are defined with respect to what catagory the other jet is in, so combinations like L2JetHigh-UnJP1 are not allowed by construction. This eliminates a further 12 combinations. If we don't care which jet is which (ie L2JetHigh-JP1Lo = JP1Lo-L2JetHigh) we can eliminate 6 more combinations by symmetry. Finally, JP1Hi is defined in such a way that if it is fired, L2JetHigh must not have fired, so the combination L2JetHigh-JP1Hi is forbidden. This leaves 8 valid jet-jet combinations:

L2JetHigh-L2JetHigh

L2JetHigh-JP1Lo

L2JetHigh-UnL2JetHigh

JP1Lo-JP1Lo

JP1Lo-JP1Hi

JP1Lo-UnJP1Lo

JP1Hi-JP1Hi

JP1Hi-UnJP1Hi

 

Figure 3: This figure shows the dijet invariant mass spectra for the 8 valid trigger combinations described above. The blue curves are data and the red curves are simulation.

 

A PDF containing data / simu comparisons for several other quantities as well as showing the comparisons for the Lo pt jet can be found here.

 

Comments:

• The jet pt data / simu agreement looks good but the agreement is not so good in the endcap region of the eta spectra (although it looks much better than it did before the 7.5% gain drop, see fig 4 here). In future versions, I will include barrel and endcap jet classifications so we can see if the data / simu disagreement has an effect on other jet quantities such as pt.

• I am still thinking about the 'completeness' of my trigger scheme, ie are there a class of jets that will be excluded from my analysis by the way I have defined the various catagories?

 

The point was made that the data / simu disagreement in the eta spectra in the endcap region may be due to a missmatch in z-vertex distributions between data and simulation. To see if this is a contributing factor, I plotted the z-vertex distributions in data and simulation for each trigger and then applied a correction to the simulation.

 

Figure 4: This figure shows the z-vertex distribution for data (Blue) and simulation (Red) for each trigger catagory. The top six pannels show the spectra and the bottom six pannels show the data / simulation ratio for each trigger along with a 4th order polynomial fit which is used in the subsequent reweighting.

 

Using the fits above, I re-ran the simulation sample with a z-vertex reweighting.

 

Figure 5: This figure shows the z-vertex distributions for data (Blue) and simulation (Red) for each trigger catagory after the z-vertex reweighting was applied.

 

Figure 6: This figure shows the jet eta spectra and ratio for each trigger catagory before the z-vertex weighting was applied (Blue) and after the z-vertex weighting was applied (Red). The top six pannels show the spectra and the bottom six pannels show the ratio without reweighting / with reweighting.

 

As can be seen in figure 6, doing the z-vertex reweighting changes the eta spectra slightly, but not nearly enough to account for the difference seen between data and simulation in figure 2.

 

 

Another issue I was concerned about was the 'completeness' of my trigger scheme. I was worried that the way my trigger catagories were set up would exclude certain classes of jets which should be included in the analysis. One case I was worried about were jets passing all the L2JetHigh conditions besides the jet pt > 8.4 condition. If this jet did not also satisfy JP1Lo, it would be lost. To test if this occurs, I changed the L2JetHigh pt requirement to be: jet pt < 8.4 and I required that the L0 JP1 trigger did not fire. This should select those jets that would be rejected in the original scheme.

 

Figure 7: Pt spectrum of jets which would have passed the original L2JetHigh conditions except for jet pt > 8.4 and did not pass the JP1Lo trigger. This is for one run only.

 

So it appears that the way my trigger catagories are set up will exclude some valid jets. Maybe I can eliminate the pt requirement and divide things by: only fire L2JetHigh, fire L2JetHigh and JP1, and fire only JP1, and the corresponding Untriggered catagories?

 

Summary of Discussion on List as of 8/14/12:

• Data / Simu disagreement of eta spectra in Endcap region doesn't appear to be caused by Z-Vertex mismatch and is likely not caused by L2 threshold or pedestal mismatches.
• Trigger scheme as implemented will cut out some jets, but these are caused by either random accept or the prescaling of the triggers. These jets are not analyzed because the systematics they introduce are too great.

 

 

"Homework" from Jet Meeting on 8/14/12:

At the jet meeting on 8/14/12, there was more discussion about the data / simu disagreement in the eta spectra in the endcap region. There were three suggestions of things too look at to get a better idea of what could be causing the discrepency:

• Make seperate data / simu comparisons for the Random, Monojet, and Dijet components of the L2Jet filter
• Look at 5-point branch
• Use exact same runs in data and simulation

 

The result of the first investigation can be seen in figure 8 below. I used the same definition of the L2JetHigh category presented above except instead of having the L2Jet filter condition of (Random || Mono || Di), I now look at each component seperately, ie Random = (Random && !Mono && !Di), Mono = (!Random && Mono && !Di) and Di = (!Random && !Mono && Di).

 

Figure 8: This figure shows the jet Eta spectra (Blue=Data, Red=Simu) and data / simu ratio for the Random, Monojet, and Dijet components of the L2JetHigh trigger.

 

The random category has too few statistics in the simulation to make any conclusions, but it looks like the mono and di categories have relatively similar behavior. It doesn't appear that one component of the L2Jet filter is driving the data / simulation disagreement.

 

I have also looked at the 5 TPC hit point branch to see if that altered the data / simulation comparison at all. The idea was that maybe the fall-off of the tracking efficiency was different between data and simulation causing more tracks to be present in the endcap region for simulation and thus creating more jets above threshold. One trigger stood out as having a large difference between data and simulation: UnL2JetHigh for the low pt jet.

 

Figure 9: This figure shows the jet eta spectra of the low pt jet.

 

For the 12 point branch in the UnL2JetHigh category, the simulation undercuts the data whereas in the 5 point branch, the data and simulation are in much better agreement.

 

I have compiled pdfs of all the plots for the 12 point branch and the 5 point branch.

 

 

Figure 10: This figure shows the L2JetHigh jet eta spectra for all jets (left column) and for simu jets which have more than one track (right column). There is no track condition on the data jets.

 

Figure 11: I have also made the above plot for jets which contain no endcap towers and jets which do contain endcap towers.