Title: Basic D Data Analysis Tutorial
1Basic DØ Data Analysis Tutorial
Frédéric Déliot CEA-Saclay
Vancouver Workshop 13 June 2005
2This Tutorial
- DØ releases and data model
- SAM
- DØ data pass and data tiers
- Which release to use for analysis ?
- DØ framework, packages and RCP
- compile and run a DØ executable
- produce a common analysis format root-tree
- compute luminosity
Other tutorials during the Workshop
Tuesday 4PM Trigsim and Rate Tool Tutorial (B.
Vachon, S. Lammers) Thursday 4PM Common
Analysis Format (CAF) Tutorial (R. Hauser)
3How DØ Reconstructed Data Looks Like ?
- DØ Event Data Model (EDM)
- representation of events in memory
- the smallest unit of an event is a chunk
(zebra bank) - an event a list of chunks from different types
(JetChunk, MuonParticleChunk, L3TriggerChunk,
HistoryChunk ) - a chunk 1 object (MissingET) or a list of
objects (vector of Jet, vector of MuonParticle, )
- packed summary of the reconstruction program
ouput - (micro-DST)
- ThumbnailChunk (all infos about objects, packed
format for JetChunk MuonParticleChunk, ) - TMBTriggerChunk (L3 trigger informations)
- HistoryChunk (event processing history),
- will have to unpack it (recreate JetChunk,
MuonParticleChunk, from ThumbnailChunk) before
using it
4How DØ Reconstructed Data Can/Will Also Looks
Like ?
- common analysis root-based format
- just finished to be developed
- derived from thumbnails
See Reiners tutorial for how to use it
5DØ Data Pass and Data Tiers (for data ? p17)
6Fixing / Skimming / Certifying
- to fix/correct/apply new calibration (mainly for
calorimeter) - using thumbnail, on the full dataset, create
fixed thumbnail - not a full reconstruction
- in the past pass1 and pass2 on p14 data
- in the future hadronic calibration on p17 data
- isolate interesting events, exclusive
- on thumbnail, create (smaller) skimmed
thumbnails - physical skims (files) logical skims (tags in
the event)
- apply standard corrections/cuts developed by the
ID groups to define good objects suitable for
most of analyses (Good EM, jets, muon, JES
corrections, ) - on thumbnail before your analysis
code/root-tuple maker - call for d0correct
7Where are the data ?In SAM !
- Sequential data Access via Metadata
- DØ (and CDF) data handling system
- store and retrieve files (from cache disks or
tapes) using metadata
- description of a file stored in SAM
- some informations in the metadata are used for
SAM query
- an abstract definition of your data
specification of metadata that will be translated
into a list of files - need to use a dataset to run an executable on
files in SAM
8DØ Releases
- a complete set of DØ software organized in CVS
packages - CVS is the DØ code management system
- development release (new software) tXX.YY.ZZ
t05.06.00 - production release (official processing)
pXX.YY.ZZ p17.03.03
- setup D0RunII pXX.YY.ZZ (non optimal build)
- setup D0RunII pXX.YY.ZZ O SRT_QUALmaxopt
(optimal build) - using of not shared libraries
- gmake all (no shared libraries)
- gmake all LINK_SHARED yes (shared libraries
created)
- Where are the releases located ?
9DØ Packages
- a group of closely related C classes
- specific versions managed by CVS
- standard structure header files directory,
source files directory, run parameters directory
- run control parameter files text files with
parameters to control what the package will do - RCPs stored in a database
- RCP of the package executable that controls what
the executable should do when it is launched - the string Packages line the list of action
to do for each event - RCP alias ltPackageName PackageRCPgt
10DØ Luminosity
- the smallest unit of time for luminosity is the
luminosity block (LBN luminosity block number),
60 s - status assigned to each LBN based on the DAQ
status (good/bad) - only data from good LBN can be properly
normalized
- luminosity infos are provided only for a
specific (set of) trigger(s) - for analysis, usually need the reconstructed
luminosity
- needed inputs to compute luminosity
- list of files you process
- list of raw data parent files (from SAM)
- your specific list of bad Run and bad LBN (that
will be added to the DAQ bad list) - you have to cut out the corresponding bad LBN in
your analysis code