Neural Network selection for the 4q channel - ep-alephsec

apricotpigletΤεχνίτη Νοημοσύνη και Ρομποτική

19 Οκτ 2013 (πριν από 3 χρόνια και 1 μήνα)

63 εμφανίσεις

27 April 02

Ann Moutoussi, CERN

Jet Properties, Data
-
MC with all Aleph Data

Jet Mass

Angles


@ Lep1 Z


Lep2 Z, W & Z
g


Aims:



to investigate general Data
-
Mc agreement



to check

consistency

of ‘picture’ between different
data samples



to find a way to

extract correction factors and/or

systematic shifts

from control samples (eg Z’s)
and apply on W’s.


The different Samples:
Z’s @ LEP1, LEP2


LEP1 Z

:
1994 Data and Latest Jetset MC


Reconstruction: All EFLOW, REPG


Selection: Franco’s Thrust>0.8, Ech>10


~500K



Lep2 Z :

e.g 1998 Data and Jetset



Reconstruction: All EFLOW


Selection: Franco’s


~100K per year




The different Samples:
W’s and Z
g

All LEP2 data & New MC: Koralw3, KK2F, ZZ, Pz (no Wev,
gg
)


W’s

: All EFLOW


4q
:
~3.5K


NN14, trainings ala CN, cut=0.3


Typical selection
-
to ensure good pairing: Standard Kinematic Fit, Pairing,
window and final selection according to 4C
-
Rescaled Mass (60
-
86, 74
-
86)
(I.e identical to ‘traditional’ measurement)


2q
:
~3.5K



Selections ala CN


Loose further selection
-
to avoid Wev,
gg
, Mhadronic>60GeV
(I.e not identical
to mass measurement where tighter windows on fitted properties are imposed)


Z
g
:
Reconstruction and basic Selection: Eugeni:
~26K


No LCAL, SICAL &
No ‘bad’ tracks, i.e V0 s


Mvis>50, Locking of
g



Loose window
-
to avoid
gg
,
120>Mhad>60 (Standard cut on x: 0.6
-
0.88)



Comment on Jet Mass at Z’s (1)


Previous meeting different numbers (~20
-
~100MeV)


Different samples, (W’s, Z’s, Z
g
) or
Same samples
-

different Selections

e.g
Thrust cut on Z events
:


Jet Mass and Data
-
MC agreement vs Thrust Cut




Jet Mass

Data
-
Mc

With JetC

1)
Jet Mass and Data
-
MC

depend on Thrust


2)
JetC: difference ~50MeV

No JetC

Comments on Jet Mass at Z’s (2)


Which Thrust selection more appropriate for W’s?

Z
:
T>0.8

Z
:
T>0.95

W: 2Q

Z
g

No selection identical, but high Thrust less similar.
For remaining: keep 0.8


Comment on Jet Mass at Z’s (3)


But.. Still Data
-
MC less than 100 MeV

Jet Mass and Data
-
MC agreement vs Thrust Cut


with & without
REPG
:
g

calibration like








LEP2



Use REPG etc


with LEP1

Jet Mass

Data
-
Mc

With JetC

(same)

Difference ~20
-
40MeV

No Jetc

Jet Mass Variation:
Data
-
MC agreement vs Pcut

Z:



1994 (~500K)




1998 (~100K)

Z
g

4Q

2Q

Different EFLOW:Track Selection..

(e.g no V0’s)

Compatible behavior??

What could make it different ??

Charged Jet Mass (no JetC)

Jet Mass Variation:
Data
-
MC agreement vs Pcut

(Different EFLOW:Track Selection

+ No LCAL/SICAL)

Jet Mass (With JetC)

Z:



1994 (~500K)




1998 (~100K)

Z
g

4Q

2Q

Scale..

Any other dependence of Jet Mass and Data
-
MC?


Data
-
MC Jet Mass vs Cos(
Q
) (Standard EFLOW)


Z :1994

Z
g

No JetC

With JetC

So…how compatible are Z’s,W’s & Z
g
’s ?


Z: Lep1 similar trends to Lep2. Maybe larger shifts at Lep2?

(15

10

vs 50

25
)


Combine all LEP2 Z (1998+1999+2000)


If we trust Z’s…
Systematic at Pcut=0 and Pcut~1: same size
, opposite
sign…



4Q: similar trends & Shifts within 1
s
to Lep2 Z’s.(70

55
)


2Q: larger shifts, but within 2
s

to Lep2 Z’s. (210

90
)


+ Caution
: Selection sensitive (to be checked/use Mass analysis Windows)


`Forgot’ some bkg?


Any chance for splitting by channel?


Z
g

~similar trend.



Check effect of different reconstruction?



Some Dependence with
Q
? (use full LEP1 data?)


Basically compatible….Propagate shifts.. With care!



Angular Biasses… at W’s and Z
g
’s
...


As always use different Jet components as a tool:





Jet direction


Neutral* Objects Vector


Charged Objects Vector

Can Measure
Q
(charged
-
charged):
Q
ch


Q
(all
-
all) :
Q
a


Q
(neutral
-
neutral) :
Q
n


Q
(all
-
charged) :
Q
ach


Q
(all
-
neutral) :
Q
an

and the differences
(
Q
ch
-

Q
a),

(
Q
ch
-

Q
n)

etc for Data and MC

and construct double difference eg:


[(
Q
ch
-

Q
a)]
Data
-

[(
Q
ch
-

Q
a)]
MC

* in following Neutral=type 4+type5 Eflow

Angular Biasses… 1)at W’s...

MC Charged
Q
ch

Data Charged
Q
ch

MC all
Q
all

Data all
Q
all

Pcut

Pcut

2Q

4Q



All


Ch.

Angular Biasses… at W’s...
[(
Q
ch
-

Q
a)]

2Q

4Q

As

Q

depends on Mw look at differences:

[(
Q
ch
-

Q
a)]
Data
,
[(
Q
ch
-

Q
a)]
MC
[(
Q
ch
-

Q
n)]
Data,
[(
Q
ch
-
Q
n)]
MC

4Q

2Q

(all EFLOW)

Points Data

Histo MC


Angular Biasses… at W’s…Double Differences

4Q

[(
Q
ch
-

Q
a)]
Data
-

[(
Q
ch
-

Q
a)]
MC
[(
Q
ch
-

Q
n)]
Data
-

[(
Q
ch
-
Q
n)]
MC

Mean Difference vs Pcut:

4Q

2Q

2Q

~
1 mrad

~
-
2 mrad

~
-
4 mrad

~
-
12 mrad

Angular Biasses… Summary at W’s...

MC Charged
Q
ch

Data Charged
Q
ch

MC all
Q
all

Data all
Q
all


Going from All to Charged Di
-
jet Angles get smaller for

both channels, Data&Mc




but.. In Data they get
Smallerer

than in MC.. (~3mrad)



Charged Only measurement will be smaller…


(especially true for semileptons.. Check selections etc..)


Biass seems flat & does not depend on Pcut.



But which is
-
more
-

correct.. Charged or All/Neutral?

Angular Biasses… 2)at Z
g
’s...

MC Charged

Data Charged

MC all

Data all

Pcut

Pcut

2Q

Z
g

1)
Not same ordering!

Is it the different

Reconstruction?





All


Ch.

Angular Biasses… 2)at Z
g
’s...

MC Charged

Data Charged

MC all

Data all

Pcut



Both Components shifted…


towards lower EcM.



Similar results for charged/All
analysis…
(to be checked again?)



Can the tracks be wrong!?!?

Angular Biasses… at Z
g
’s…Double Differences

4Q

[(
Q
ch
-

Q
a)]
Data
-

[(
Q
ch
-

Q
a)]
MC
[(
Q
ch
-

Q
n)]
Data
-

[(
Q
ch
-
Q
n)]
MC

Mean Difference vs Pcut:

4Q

~
2 mrad

~
-
6 mrad

Pcut

Pcut

Angular Biasses… Summary...

W:


Cannot tell if Charged/Neutral more wrong.


Can use difference as an estimate of angular biass.



Semileptonics seem more affected…


again to check selections etc.



Difference charged
-
neutral to be propagated.



Z
g
:

If we believe Elep:



Could be used to tell if charged/neutral better..



Ordering of angles not as in W’s.
(Charged angle larger than EFLOW angle)



Check reconstruction



Check with Paolo angular differences