Lhep: 1 revision imported

2015-03-18T07:49:23Z

1 revision imported

← Older revision	Revision as of 09:49, 18 March 2015
(No difference)

Lhep at 10:19, 5 May 2010

2010-05-05T10:19:22Z

New page

== Talks ==

Walk-through presentation (MiniBat, 16/11/2009) [http://indico.cern.ch/getFile.py/access?sessionId=1&resId=0&materialId=0&confId=71656]

Bern Plans in 1 lepton channel (Susy ETmiss meeting, 14/01/2010) [[http://www.lhep.unibe.ch/lhepwiki/images/4/40/BernPlans.pdf]]

Analysis status (Bat, 25/2/2010) [[http://www.lhep.unibe.ch/lhepwiki/images/1/18/Phase1_250210.pdf]]

JE trigger studies (Susy Trigger Meeting, 23/3/2010) [http://indico.cern.ch/conferenceDisplay.py?confId=87452]

JE trigger studies (Jet Trigger Signature Group Meeting, 03/05/2010) [http://indico.cern.ch/getFile.py/access?contribId=16&resId=1&materialId=slides&confId=93128]

== Comments Michele to the walk-through Nov 09 ==

----
Where are you in the plans on page 24 ?

----
Are you working in release 15 ?
----
Where in SVN is your analysis code ?

----
A first step is to get the signal and BG samples (list on p18 ?).
Repeat yourself the steps as in page 7.

----
For all the bg samples make a table with cross sections, number of
events before selection, after selection, ...

----
I agree to go for the 1-lepton selection as a first go.
Think about doing e or mu, or both ? (see also your slide 13).
This depends on who else will do a benchmark base
exclusion/search analysis to be completed by next summer. Please find
out. There is also the possibility to add b-tagging requirements.
Is that an option ?

----
About the limit statistics: make a proposal of what you will use.
CLs ? But do not spend too much time on this at first.
This you can redo later if you have all the numbers. Applying a
more sophisticated method. I would go with frequentist first
(heavy-heartedly...)

----
Talk to Tobias about limits in SU4. Do you get the same estimates ?

----
Event selection. Will you start with the CSC 14 TeV default ?
Or do you have a better proposal ?

----
It may be a good idea to simulate the analysis in cosmic data ?
We do not expect to find anything, but it would be a nice exercise.
Or on the top-mix samples ? Or on the first runs with 900 GeV
collisions ? To get a _real_ dataset for the analysis will take us to
late spring at least.

----
Trigger strategy for first data in the SUSY Group.
Find out from Teresa what the strategy is. What triggers to use ?
Are the jet tiggers filtering at HLT ?

----
Since the trigger is a significant part of your thesis: propose a
trigger to be used. The "full thing" ! Trigger, L1, L2, HLT (?), rates
(minbias), efficiencies (signal) estimated from turn-on (Tobias) and
simulation, overlaps, comparison to other triggers. Validation in
first data. I guess this could be a 3J_XXX. Focus on one first. Once
you have an idea, keep Teresa and George informed. Talk to Tobias
about code and possible studies of turn-ons on QCD samples.
For the selection efficiencies on signal, remember that they are with
respect to the offline selection.

----
How do you plan to apply the triggers in MC ? Thrhw away some events ?
Weights ?

----
A study of the HLT in PT makes sense for a trigger study. But you
don't care for the physics analysis. L1 alone, if possible for the
rate and efficiency, is easiest. Likely possible for L1_3J10 or
L1_3J20.

----
I would not plan on using triggers that may be prescaled, to start
with. The prescale factor can change dramatically and in a non
predictable way in first running. It will be more difficult to get
the effective luminosiy.

----
How do you plan to apply data quality and trigger efficiency ?
As an efficiency for your signal, or do you think to reduce to an
effective luminosity ? Both are very viable. Think about it.
Can you estimate if the 3J_10 could be safe to run unprescaled ?
One issue is noise in the calorimeter. This could be evaluated in
cosmics running (check the 3J_10 rate, or better search for 3J_10
triggers...). Also could be done from first collisions on a minimum
bias sample anc check for 3J_10 trigger (also here not many... if
everything is as expected).

----
The object definition. What are the definitions for first data in the
SUSY group ? Do you plan to take the ones from the Jet/EM/Muon
walk-thoughs ? Is there a study for 7 TeV ?
I agree you should take whatever is available and not spend time in
studies for now. But you need a starting point. Talk to Tommaso Lari ?
Once you have an analysis, you can take another object definition and
compare. And take the difference as a systematics.

----
One point I am sensible... The separation of object definition and
event selection. For example jets: you will require jet-pt>50,
nevertheless also a calorimeter object of 15 GeV is a jet.
This affect the n-jet quantity in the event.

----
Make sure you event selection is exclusive (orthogonal) to possible
other selections in other analyses. This is to make sure we could
combine the results in an easier way.

----
Tobias is following the Freiburg selection. It may be a good idea to
start from that too ?

----
Find out more about the delta-Phi(jet,met) cut.
Is this to reject QCD ?

----
Start cut based. Then implement a likelihood if you have time.

----
You will need to use Met. But the values you look at are rather high
(> 50 GeV). Please provide a Met resolution plot. Either expected, or
from cosmics, or from first collisions. Feel free to ask Joel to help you.

----
Look at the estimates / numbers that the data drive BG groups
provide. Do you feel that you can use them ? Just take the numbers ?
I believe for QCD we will have to go the data driven way.
W+jet and ttbar are better under control.
Who can provide you with a reasonable QCD (multijet sample) from
data ?

----
Who in the SUSY group will provide a good-runs-list ?
Someone defined the the flags to be used ?
When will this happen ? How often ? Daily ?
Will the luminosity be provided ?

----
How is the bookkeeping done ? How are duplicated events removed ?
How is made sure, that you processed all files corresponding to an LBN
list ?

----
A lumiblock based DQ is applied. Also an event based ? (noise)

----
Are there plans to overlay minimum bias events from data to the MC
(for noise, multiple interactions, etc...) ?

----
It would be nice to use some ensemble testing techniques.
Which means doing many pseudo-experiments in MC to estimate the
fluctuations / limits. this could be done at 1 fb-1 and then the
expectations scaled down to 10-100 pb-1.

Andreas - Revision history

Lhep: 1 revision imported

Lhep at 10:19, 5 May 2010