Bruce Knuteson
Global Analysis of High-p
T
Data
The problem
The solution
Vista
Sleuth
Surprise!
Bard
Quaero
TurboSim
LHC New Physics Signatures Workshop, University of Michigan, Jan 6 2008
Bruce Knuteson
2
The problem
model space is really, really big
Problem
Solution Vista Sleuth Surprise! Bard Quaero TurboSim
10
105
Bruce Knuteson
3
Z’
W’
π
T
LQ
A
Z’
W’
A
π
T
LQ
Problem Solution Vista Sleuth Surprise! Bard Quaero TurboSim
The solution: Look everywhere
Bruce Knuteson
Define physics objects
Filter events of interest
e
γ
μ
j
b
ν
τ
pp
Estimate backgrounds
Simulate detector response
tr
ue
reconstructed
(mis)Id
Fit for experimental & theoretical factors
Vista algorithm
e
μ
τγ
jb
e
0.62
2e-3 0.02 0.28
μ
0.51
τ
0.02 0.01 0.04
0.90 6e-3
γ
0.03
0.68 0.21
j
1e-4 1e-5 3e-3 3e-4
1
2e-2
b
1e-4 1e-4 1e-4 5e-5 0.65 0.35
Problem Solution Vista
4
Sleuth Surprise! Bard Quaero TurboSim
hep-ex/0402029
hep-ex/0504041
arXiv:0712.1311
arXiv:0712.2534
Bruce Knuteson
Vista correction factors
(global fit)
CDF Run II (927 pb
-1
)
Bruce Knuteson
6
The Vista result
Georgios Choudalakis
MIT
Conor Henderson
MIT
Ray Culbertson
FNAL
Problem Solution Vista
Sleuth Surprise! Bard Quaero TurboSim
arXiv:0712.1311 (submitted to PRD 10-Dec-2007)
arXiv:0712.2534 (submitted to PRL 15-Dec-2007)
Bruce Knuteson
7
Vista
output
(stat. uncertainty only)
CDF Run II (927 pb
-1
)
Problem Solution Vista
Sleuth Surprise! Bard Quaero TurboSim
Bruce Knuteson
8
Vista
output
CDF Run II (927 pb
-1
)
Problem Solution Vista
Sleuth Surprise! Bard Quaero TurboSim
Bruce Knuteson
9
Vista
final state normalizations
Bruce Knuteson
data < SM
SM = data
SM < data
CDF Run II (927 pb
-1
)
Problem Solution Vista
Sleuth Surprise! Bard Quaero TurboSim
σ
Bruce Knuteson
10
Vista
kinematic shapes
Bruce Knuteson
CDF Run II (927 pb
-1
)
agreement disagreement
σ
Problem Solution Vista
Sleuth Surprise! Bard Quaero TurboSim
Bruce Knuteson
1.
Statistical fluctuation
2.
Detector effect
3.
Poor prediction
4.
Plausible interpretation
Problem Solution Vista
Sleuth Surprise! Bard Quaero TurboSim
Problem Solution Vista
Sleuth Surprise! Bard Quaero TurboSim
Bruce Knuteson
12
Bruce Knuteson
r
z
r
z
1
2
3
1
2
3
Sample discrepant distribution (parton showering suspected)
Problem Solution Vista
Sleuth Surprise! Bard Quaero TurboSim
Bruce Knuteson
13
Related discrepant distribution
BrBrucuce e KnKnuutteesoson n
Problem Solution Vista
Sleuth Surprise! Bard Quaero TurboSim
Bruce Knuteson
14
Sleuth
Assumptions:
1. Exclusive final state
2. Large ∑p
T
3. An excess
0001001
present
(prediction) d(hep-ph)
Rigorously compute the trials factor
associated with looking everywhere
DØ Run I
Phys.Rev.D 62:092004,2000
Phys.Rev.D 64:012004,2001
Phys.Rev.Lett.
86:3712,2001
H1 General Search
Phys.Lett.B 602:14-30,2004
arXiv:0705.3721 (summer 2007)
CDF Run II
Back to top
arXiv:0712.1311
(submitted to PRD)
arXiv:0712.2534
(submitted to PRL)
∑p
T
statistic
a quasi-model-independent search strategy for new physics
Problem Solution Vista Sleuth Surprise! Bard Quaero TurboSim
Bruce Knuteson
15
Find the most
interesting tail
Produce pseudo
experiments to
assess
significance
411
Problem Solution Vista Sleuth Surprise! Bard Quaero TurboSim
CDF Run II (927 pb
-1
)
Bruce Knuteson
16
Jet (j)
Jet (j)
Sleuth
Sensitivity
Q:
Would Sleuth have
found the top quark?
A: Yes.
Expected Run II discovery
luminosity ~80 pb
-1
(Run I discovery 67 pb
-1
)
CDF Run II (927 pb
-1
)
Bruce Knuteson
17
Bruce Knuteson
top quarks removed from the background estimate
discovery threshold
~
~
Problem Solution Vista Sleuth Surprise! Bard Quaero TurboSim
CDF Run II (927 pb
-1
)
Bruce Knuteson
18
Sleuth
targeted search
σ
discovery
(pb)
better
worse
Sample comparison of
Sleuth
to targeted searches
Problem Solution Vista Sleuth Surprise! Bard Quaero TurboSim
CDF Run II (927 pb
-1
)
For models
satisfying the
assumptions on
which Sleuth is
based, Sleuth is
comparable in
sensitivity to a
targeted search
Bruce Knuteson
19
The Sleuth result
∑p
T
statistic
Problem Solution Vista Sleuth Surprise! Bard Quaero TurboSim
arXiv:0712.1311 (submitted to PRD)
arXiv:0712.2534 (submitted to PRL)
Bruce Knuteson
20
46% of pseudo experiments are
expected to be as interesting
Sleuth finds no significant excess
in CDF Run II high-p
T
data
This does not
prove there is no
new physics present
Sleuth@CDFII
Back to top
result
(top 5)
Back to top
fraction of pseudo experiments in this
final state as interesting as CDF data
Back to top
fraction of pseudo experiments in any
final state as interesting as CDF data
Problem Solution Vista Sleuth Surprise! Bard Quaero TurboSim
Back to top
CDF Run II (927 pb
-1
)
BrBrucuce e KnKnuutteesoson n
Problem Solution Vista Sleuth Surprise! Bard Quaero TurboSim
CDF Run II (927 pb
-1
)
Bruce Knuteson
Problem Solution Vista Sleuth Surprise! Bard Quaero TurboSim
22
Scientific Merit = how much you learn
= how surprised you are at the result
= surprisal
= -log
10
(p)
Bruce Knuteson
Problem Solution Vista Sleuth Surprise! Bard Quaero TurboSim
23
Table II, pg 8
† Hypothetical future result
Bruce Knuteson
Problem Solution Vista Sleuth Surprise! Bard Quaero TurboSim
24
Nobel Prize
contenders
Table II, pg 8
† Hypothetical future result
Bard
hep-ph/0602101
Knuteson, Mrenna
Problem Solution Vista Sleuth Surprise! Bard
Quaero TurboSim
Bruce Knuteson
26
Standard
Model
Data
Problem Solution Vista Sleuth Surprise! Bard
Quaero TurboSim
Bruce Knuteson
27
Bard
Problem Solution Vista Sleuth Surprise! Bard
Quaero TurboSim
Bruce Knuteson
28
Bard
Problem Solution Vista Sleuth Surprise! Bard
Quaero TurboSim
Bruce Knuteson
29
Bard
Problem Solution Vista Sleuth Surprise! Bard
Quaero TurboSim
Bruce Knuteson
30
Bard
Problem Solution Vista Sleuth Surprise! Bard
Quaero TurboSim
Bruce Knuteson
31
Problem Solution Vista Sleuth Surprise! Bard
Quaero TurboSim
Bard stories
Bruce Knuteson
Quaero@D0RunI
Back to top
DØ Collaboration
Phys.Rev.Lett.87:231801,2001
Back to top
Quaero@H1
Back to top
S. Caron, B. Knuteson
Eur.Phys.J.C53:167-175,2008
Problem Solution Vista Sleuth Surprise! Bard Quaero
32
TurboSim
Bruce Knuteson
TurboSim
Back to top
A fast detector
simulation that
tunes itself
to any
Back to top
experiment’s
Back to top
detailed detector
Back to top
simulation
Full simulation
100 seconds
TurboSim
0.01 seconds
Bruce Knuteson
MIT
MIT
MIT
Problem Solution Vista Sleuth Surprise! Bard Quaero TurboSim
Bruce Knuteson
34
If a core group of 4
people pursue Vista,
it is an endgame
If a core group of 24
people pursue Vista, it
is an opening gambit
Problem
Solution Vista Sleuth Surprise! Bard Quaero TurboSim
Bruce Knuteson
Global Analysis of High-p
T
Data
LHC New Physics Signatures Workshop, University of Michigan, Jan 6 2008
The problem
The solution
Vista
Sleuth
Surprise!
Bard
Quaero
TurboSim
Summary
Back to top
