2006 Inclusive Jet pT Shifts and Trigger and Reconstruction Bias - TAKE 2


To determine the final pT shifts the monte carlo set was divided into four sub-samples based on the changing
trigger thresholds during the running period.  From Carl's post to the spin mail server
(http://www.star.bnl.gov/HyperNews-star/protected/get/starspin/3435.html), we see that there are four main time windows:


Run Range: 7131043 - 7132027
    - JP1: ID=137221  Thr=58
    - HTTP: ID=5
         HT1 thr=16, TP1 thr=20, tower seed = 3.00, patch = 4.79
    - HT2: Not included in physics analysis
Run Range: 7132062 - 7133051
    - JP1: ID=137222  Thr=60
    - HTTP: ID=5  --- prescaled at 2 for a fraction of range
         HT1 thr=18, TP1 thr=20, tower seed = 3.00, patch = 4.79
    - HT2: ID=137213 Thr=24
Run Range: 7133052 - 7135028
    - JP1: ID=137222  Thr=60
    - HTTP: ID=5
         HT1 thr=16, TP1 thr=19, tower seed = 3.80, patch = 5.10
    - HT2: ID=137213 Thr=24
Run Range: 7135067 - 7156028
    - JP1: ID=137222  Thr=60
    - HTTP: ID=137611
         HT1 thr=16, TP1 thr=19, tower seed = 3.80, patch = 5.20
    - HT2: ID=137213 Thr=24
The fraction of the total data taken from each period using the final Long-II run list:
- 1st time window: ~1.5%
- 2nd time window: ~9.8%
- 3rd time window: ~10.9%
- 4th time window: ~77.8%
Our final ALL figures use the OR-sum of the three triggers over the whole run range.  The goal of this study is to match the data OR-sum with the OR-sum for simulation - this means that the trigger overlaps must match.  
One further complication is that there is a prescale for the HTTP trigger for a portion of the second time window.  This was simulated by applying a 0.566 weight to jets in time window 2 passing HTTP but not HT2 or JP1 (so explicitly, HTTP && !(HT2||JP1)).  0.566 is used instead of 0.5 as the weight because a prescale of 2 was not present the entire time window.  Below shows the trigger overlaps comparison between data and simulation - they match well.
To calculate the pT shifts, I numerically calculated the average Pythia pT value for each Geant pT bin.  Posted below are the the shift values obtained for each trigger seperately and the OR-sum trigger combination.  Also shown are the OR-sum values with a fit to demonstrate that the uncertainties are reasonable and the values follow a trend that is characterized with only a few parameters.
Jet/Event Cuts Used:
 - zVertex != 0
 - Geant ETA >-0.7, <0.9
 - Geant jet pT >= 5.0
 - Geant jet neutral Energy Fraction <0.92
 - Murad's electron-like Jet Cut: http://cyclotron.tamu.edu/star/2005n06Jets/eJets/
 - Software Trigger Requirements
 - For the pT-shift analysis specifically (i.e., not the case in trigger/recon bias below), only associated Geant+Pythia jets are included in the calculation.  A cut on dR=sqrt(dPhi*dPhi + dEta*dEta) < 0.3 is used: see this figure.
The number of events per pythia sample used is shown in this file as is the calculated partial cross sections used to weight the various sample relative to each other.



 Posted here are the values to be used in our figures.




 To estimate the bias from our trigger and jet reconstruction we shift the Geant asymmetries to these pythia mean pT values and compare to raw Pythia asymmetries.  For 2006 we have the ability to look at many theoretical parameterizations of delta-G: 11 GRSV models, 3 GS models, 3 LSS models, 3 AAC models, 2 BB models, and 2 DNS models.  See my previous drupal post for more details about these various theoretical modes.  The below figures show a comparison between three curves for each of these models: 1- Pythia A_LL values (blue point, lines), 2- Geant un-shifted A_LL values (black), and 3- Geant shifted A_LL values (red).  The OR-sum trigger combination is the only displayed below.



The difference pythia ALL minus shifted-Geant ALL is displayed below for all the parameterizations used above:



Posted here are the difference for each pT bin and parameterizations.  Also, to be thorough, posted here are the geant asymmetry values for each parameterization and pT bin.