1 / 7

Analysis model s in the top physics group

Analysis model s in the top physics group. A few remarks … What people are doing now Where we are going - Top reconstruction framework Input and output formats General remarks N.B. Not official top group ‘policy’ - personal thoughts

may
Télécharger la présentation

Analysis model s in the top physics group

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Analysis models in the top physics group • A few remarks … • What people are doing now • Where we are going - Top reconstruction framework • Input and output formats • General remarks • N.B. Not official top group ‘policy’ - personal thoughts • … from someone who was not particularly happy with the AMF report! Richard Hawkings CAT physics meeting, 3/4/09 Richard Hawkings

  2. Creative anarchy • March 2009 poll initiated by Nabil Ghodbane • Categories not discussed in advance, and not exclusive … 116 responses • ‘Traditional methods’ dominate (direct AOD analysis, making private ntuples) • SingleTop group D2PD/ntuple also popular (not only with single top group) • Dump of selected AOD object info, plus overlap removal and selection • Some use of ARA and TopPhysD3PD (successor to TopView ntuples) • Less than 50% use a ‘centrally-produced’ DnPD, rest are more ‘hands on’ Richard Hawkings

  3. Common top reconstruction code / framework • Some negative aspects of the ‘creative anarchy’ • Very difficult to implement, change and cross-check common object selections • Very difficult for groups to work together / validate each other’s work • Same mistakes (e.g. in truth analysis) are made again and again • Moving forward with common top reconstruction code (from TopReco group) • A pseudo-physics subgroup to provide common forum/tools/code/selections for all ATLAS top analyses - unique in the physics groups • TopInputs infrastructure - now ready (see AtlasProtected/TopInputs wiki) • Provide common object selections and overlap removal tools • Using ISelector interfaces ‘approved’ by PAT but largely being implemented by TopReco • Implementing CSC-style selections, updated to rel 14/15, as default • Instead of iterating over ElectronContainer, iterate over TopInput electrons • A bool method gives Use? Yes/No information • Link to more detailed overlap ‘reasoning’ information coming later • TopInputs has persistency - can be added to DPD files • Expect this object selection to be used by all top physics analyses • None of this is top group specific - hope it can migrate to PAT… Richard Hawkings

  4. Common top reconstruction code continued • TopEDM objects have been defined • Classes to represent reconstructed top and ttbar final states • Basically collections of ElementLinks to the constituents of the reconstructed object • Minimal additional functionality - don’t want to define things that don’t always make sense (e.g reconstructed mass of leptonic top quark decay) • Template-based approach: <lepton>+jets, <lepton><lepton> etc • Classes have persistency so can be added to DPD files • Top reconstruction algorithms are being collected/implemented in framework • E.g. ‘commissioning’ algorithm (3-jet hadronic), tt-bar likelihood and chisq fitters • These start with TopInputs and produce TopEDM objects (in containers) • ‘Decoration’ of TopEDM objects by specific reconstruction algorithms • E.g. adding masses, chisq, fit probabilities etc • Proposal is to make new classes deriving from TopEDM classes and adding information (e.g. TopElectronJetsChisq) … but the persistency is tricky • Alternatives: add <map> of extra info, egDetails-style object or use UserDataSvc • … best solution(s) not clear yet Richard Hawkings

  5. DnPDs in this model • TopInputs and TopEDM have POOL persistency • Can write D{1,2}PDs with input objects of interest (Electrons, Jets etc) + information on top object selection and reconstruction • Resulting DPDs could be analysed in Athena or ARA • Clear need also for ntuples (D3PDs) to look at the output in ROOT • + Easier for newcomers, faster to read, more intuitive • - The disadvantages of leaving the ‘framework’ • Need a way to write ntuples for input objects - various possibilities • AMA (Max Bakk already wrote AMA interface for some TopEDM objects) • SingleTopDPDMaker, EWPA or EventView • … all have modules for writing e.g. an Electron (or subset of info) to ROOT ntuple, either directly or via some intermediate transient model • Should be trivial to extend them to TopInput and TopEDM objects • … it would be desirable to have ONE officially supported way of doing this Richard Hawkings

  6. Input formats for top reconstruction • Top reconstruction code runs in Athena … • Can use any format where appropriate SG containers are available (e, mu, jet etc) • ESD, AOD, some performance DPDs which have appropriate selection • DPD_SINGLEEL (medium e, pT>15 GeV) and DPD_SINGLEMU (isol.  pT>20 GeV) • Top is a physics WG, not a performance WG • In principle, should be working from ‘standard’ objects on AOD provided by combined performance groups, no need for ESD access • But … perhaps not true in the beginning • E.g. not all potentially-interesting jet calibrations/algorithms available on AOD • E.g. may want more detailed lepton information when studying lepton efi Ztop • Performance DPDs may offer a good solution for this • Top event volumes are modest, once we understand selection basics • E.g. lepton+4 jet selection is ~50k events in 100 pb-1 • Good potential for using the TAG database and making e.g. AOD skim • Provided TAG variables correspond to the jet/lepton definitions we actually use • TopMixing sample being made available in TAGs db - first real ‘physics driven’ test Richard Hawkings

  7. Final remarks • Working towards more commonality in the top group • Common inputs and reconstruction code - Athena based • Common DnPDs, potentially POOL and ROOT-ntuple format • Tool for producing the latter is not clear yet… • Common analysis code yet to be tackled, but needed e.g. for initial top x-sec • … Easier to achieve commonality across all groups in an analysis (10-20 groups in case of top cross-section) than across all analyses going on at CERN • … Maybe less true in case of combined performance work • Gratuitous remarks • ARA is good for validation - looking at the contents of ESD/AODs etc • Personally not convinced it is a viable analysis framework, esp. for less-experienced • Already imposing ‘requirements’ on EDM; will end up re-inventing Athena - use Athena! • Lots of focus on O(10) data volume reduction: DPD streams;, 100k 10k/event • In the case of top, final analysis will be done with much smaller samples • … Keep flexibility - AOD skims, dropping containers and objects, … • But hard to define this now until we start getting to grips with the real data Richard Hawkings

More Related