Systematic analysis of frontier energy collider data

1
Systematic analysis of frontier energy collider
data
Geneva, Switzerland
Hamburg, Germany
The Game Strategy and tactics Implications
Batavia, IL
Bruce Knuteson MIT
Moriond QCD 17 March 2005
2
The first sign of new physics will come
from Heavy gauge bosons Technicolor Large
extra dimensions Leptoquarks Supersymmetry Four
th generation fermions Compositeness Something
else
7
11
21
3
36
5
6
11
3
The values of the 105 MSSM parameters are
4
Discovery with a typical search Remember that
George will guess wrong But lets say George is
very lucky and guesses close . . .
true signal
George's guess
George's modified guess
y
an "event" (pp collision)
x
5
How should George interpret this? (Will anyone
believe him?) Moral Any successful dedicated
search is bound to end up in a set of highly
sculpted cuts, since the original guess will be
wrong.
true signal
George's guess
George's modified guess
y
an "event" (pp collision)
x
6
The Game
Maximize chance for discovery before LHC turn-on
Maximize chance for discovery after LHC turn-on
7
Strategy and tactics
in brief
8
1
Vista
Model-independent search for new large cross
section physics Gross features, bulks of
distributions
9
Define physics objects
Simulate detector response
e
?
Filter events of interest
?
b
p
p
?
j
?
Fit for experimental theoretical fudge factors
(mis)Id
reconstructed
Estimate backgrounds
e ? ? ? j b
e 0.91 0.02 0.001 0.07 1e-3
? 0.87
? 0.10 0.90
? 0.81 0.19
j 1e-4 2e-6 3e-3 6e-4 1 2e-3
b 1e-3 1e-3 5e-3 8e-4 0.60 0.40
true
10
toy example
11
2
Sleuth
Quasi-model-independent search for new high-pT,
small cross section physics Excess on high-pT
tails, with rigorous computation of trials factor
12
events
Sleuth_at_DØ
H1 General Search
Phys.Lett.B 60214-30,2004 hep-ex/0408044
Phys.Rev.D 62092004,2000 hep-ex/0006011 Phys.Rev
.D 64012004,2001 hep-ex/0011067 Phys.Rev.Lett.86
3712,2001 hep-ex/0011071
13
3
Quaero
Model-dependent hypothesis testing Rigorous,
fast, transparent, robust, and automatic testing
of specific hypotheses
14
(No Transcript)
15
TurboSim
A fast simulation that tunes itself to any
experiments full simulation
v TurboSim_at_
TurboSim
Full Simulation 10 s/event
GeantSim
TurboSim 10 ms/event
16
4
Bard
Model construction Automate theorists
17
Use Quaero to determine parameters quantify
goodness of fit
Hint from Vista or Sleuth
ee- bb
? pT
Generate many possible signals
Order resulting list
18

• Legend
• published
• technically complete
• work ongoing
• ? stalled

Current scorecard
Quaero
Vista
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
Bard
?
?
?
?
?
?
?
Algorithm
no collaboration commitments
19
Implications
20
vs.
There could be a clear winner
21
Astro is ahead of HEP
22
Bio is ahead of HEP
23
PHOBOS
nucl-ex/0301017
nucl-ex/0406021
HIP is ahead of HEP
24
We are beginning to automate the scientific method
25
Summary
You cant guess right guaranteed surprise

• In light of this, a few of us are taking a
  different approach
• Gather all data
• Look at them Vista (model independent)
• Look at them again Sleuth (quasi model
  independent)
• Automate hypothesis testing Quaero (model
  dependent)
• Interpret underlying physics Bard (model
  construction)

Systematically maximize chance for discovery
Thanks to
26
Appendix
27
Bard
ee- bb
? pT