systematic errors associated with pid
DESCRIPTION
Systematic errors associated with PID. Milind V. Purohit BaBar Analysis Tools Workshop October, 2005. The PID Systematic Error Issue. The majority of BaBar analyses use some sort of particle ID Systematic errors associated with the PID efficiency are necessary - PowerPoint PPT PresentationTRANSCRIPT
Systematic errors associated with PID
Milind V. Purohit
BaBar Analysis Tools Workshop October, 2005
2Milind V. Purohit, Univ. of South Carolina
The PID Systematic Error Issue The majority of BaBar analyses use some sort
of particle ID Systematic errors associated with the PID
efficiency are necessary There is no prescribed way to obtain these The need for precise efficiencies increases
with time; e.g., upcoming CP violation studies in charm decays will need sub-1% particle ID efficiency errors.
3Milind V. Purohit, Univ. of South Carolina
What is being done today To understand better the current situation,
we can look at recent analyses. A quick scan of ~50 analyses (BAD notes) describing recent analyses (starting from the Pub Board’s 2005 summer papers list) for PID systematics information shows that systematic errors are based on:o Data-MC comparisonso Using PID weight statisticso Using PID killing vs no killingo Other methodso Unclear or no explanation
4Milind V. Purohit, Univ. of South Carolina
Summary of Some PID Systematic Error DeterminationsAvailable at http://www.slac.stanford.edu/~purohit/internal/PidSyst.htmle (%) mu (%) pi (%) K (%) p (%) BAD #s Method, Notes
0.5 1213, 824, 1213 Data vs. MC: Control Sample vs. PID group efficiencies
5 1205, 971 Data - MC comparison. Depends on D0 mass cut.
1.3 1259, 1184 Data - MC comparison.
3.5 967, 1076 Data - MC comparison.
1 1077, 664 Data - MC comparison.
0.5 0.5 1137, 824 Data - MC comparison.
0.7 - 3.8 1075, 938 PID weight statistics.
0.49 0.75 1179, 825 PID weight statistics.
2.2 2.2 1147, 323 PID weight statistics.
3.0 1159, 1013 PID weight statistics.
2 3 2 1255, 1214 PID weighting. Using PID weight statistics?
5 5 5 1187, 542 PID killing vs. No PID killing.
1.0 2.0 0.7 1105, 1032 PID killing vs. No PID killing.
3.5 1107, 1071 PID killing vs. No PID killing.
3.0 1129, 1044 PID killing vs. No PID killing.
1 1239, 1088 PID effect cancels, but accounts for various algorithms.
1.1 1243, 818Giampiero Mancinelli's study. See http://www.slac.stanford.edu/BFROOT/www/Organization/CollabMtgs/2001/detDec2001/Wed1a/giampi.pdf
2 2 2 658 "Based on data-MC comparisons."
0.5 1077, 664 Entire hadronic Mis-ID rate.
3 1135, 94 Different running periods.
5.2 1271, 1203 Unclear.
1.0 1027, 768 Unclear.
2.0 1037, 902 "Common BaBar practice."
0.2 - 2.0 0.2 - 2.0 1225, 1146 Unclear.
2.6 1243, 818 No explanation.
2 1073, 697 "[As in] similar analyses."
20 1077, 664 Arbitrary.
2.0 3.0 1161, 1091 No explanation.
1 1107, 1071 No explanation.
5Milind V. Purohit, Univ. of South Carolina
Summary of current situation Data-MC comparisons:
may simply be validating the simulation, as opposed to providing a real systematic
Use PID weight statistics:certainly a good idea, but is it sufficient?
Use PID killing vs. no killing:a variation of Data-MC comparisons
Other methods etc.:over-estimates, guess-timates, appeals to “common knowledge” and no explanation
6Milind V. Purohit, Univ. of South Carolina
What should be done and how can the PID group help? The PID group’s PID efficiencies should come
with both statistical and systematic error estimates
The best way to estimate PID systematic errors is:(Fill in the blanks here)
If we knew the preferred technique, we would work on implementing it. Your input and work is needed!
7Milind V. Purohit, Univ. of South Carolina
An example of work on PID systematics A South Carolina student, Ryan White, is
trying to address some of these issues:
Compare efficiencies obtained by different techniques:o Compare MC truth efficiency to standard PID efficiencies
and try to understand differences.o Question: are differences due impurities and differences in
samples? Compare efficiencies for kaons obtained from
different sources:o Compare kaon efficiency for kaons from D0s to kaons not
from D0s. o Question: are differences due to impurities in one or both
sources?
8Milind V. Purohit, Univ. of South Carolina
Kaon from MC Truth vs. MC as Data
9Milind V. Purohit, Univ. of South Carolina
10Milind V. Purohit, Univ. of South Carolina
2 Contribution Due to the Effect of Different Distributions with Bins
Selector (unadjusted) (bin dist.) /ndofadjusted)
K- Very Loose 172.8 2.6 1.45
K- Loose 144.7 4.9 1.20
K- Tight 137.7 6.8 1.12
K- Very Tight 131.4 6.9 1.06
K+ Very Loose 209.4 2.0 1.77
K+ Loose 191.0 3.9 1.60
K+ Tight 164.1 5.5 1.36
K+ Very Tight 176.9 5.9 1.46
11Milind V. Purohit, Univ. of South Carolina
Charge Asymmetry
12Milind V. Purohit, Univ. of South Carolina
Tracking Efficiency versus Decay Distance – Kaon Decay Mode
Decay Mode Branching Fraction
63.43%
- 21.13%
5.576%
e-e 4.87%
- 3.27%
1.73%
)()()(
unmatchedNmatchedNmatchedNeff
truthtruth
truthtracking
13Milind V. Purohit, Univ. of South Carolina
PID Efficiency as a function of decay distance for kaon decay mode
14Milind V. Purohit, Univ. of South Carolina
Kaons Interact with the Detector
15Milind V. Purohit, Univ. of South Carolina
Summary Different methods to estimate kaon systematics
have been surveyed New approaches to estimate kaon systematics
are being undertaken As questions get answered, new questions are
being raised We need input on what is needed and feedback
on whether we are headed in the right directiono Should we extend kaon studies to other particles?o Manpower is needed to do an exhaustive study of PID
systematics Analysts are encouraged to volunteer their
efforts. We can learn from their experiences with PID.