LSC-Virgo Burst Analysis Working Group

Navigation

Burst Group Home
Wiki
ExtTrig Home
CBC Group Home
LSC, LIGO, Virgo

Documents

'How-to' docs
Agendas/minutes
[2008, 2007, earlier]
Talks/Posters [pre-wiki]
Papers
Paper plans
White papers
ViewCVS

Investigations

Analysis projects
Old notebook [General, S2, S3, S4, S5]
Virgo workarea
External collabs

Review

Main review page
Telecons

Review Committee Meeting Monday 8 May 2006 09:00 PST / 12:00 EST

Minutes: Monday 8 May 2006 09:00 PST / 12:00 EST

Agenda and Contact Info

Agenda

  1. Continue the review of the S2-S3-S4 GRB-GWB search with focus on details of the analysis and tuning choices [ HTML ]
  2. Continue the review of the SGR1806-20 search with focus on details of the analysis and tuning choices [ HTML ]

Contact Info

AccuConference teleconferencing service:
   Phone: 1-800-704-9896, participant code: 038621#
   International callers ++1-404-920-6472 with same code

Minutes

  1. Continue review of GRB search.
    • Request for access to intermediate results, data, files, etc. for GRB searches. E.g., preprocessing script, parameter files, specific on- and off-source times, input files for Matlab code, output files from Matlab code, etc. Since runs were done at LLO, LHO, CIT, can make links to analysis directories.
    • Various statistical tests were considered.
      • FMR proposed t-test but this is not used since assumption that the duration of the GRB equals the duration of the GWB is not valid and the test is not applicable to the maximum of the CC stats.
      • t-test was replaced by sum-max (parametric) and rank-sum (non-parametric) tests. Sum-max is an unweighted sum: assumption is that the detector noise is stationary.
          Question
        : if varying sensitivity (parameterized by an overall scalar) were accounted for in the likelihood, would a different statistic emerge? Would the test become more powerful? Soumya will investigate, but expect that future multidetector methods will be more powerful still.
      • KS test had been considered along with binomial test. While the sum-max and rank-sum test lumps all the "foreground" results together and all the "background" results together, Isabel uses the background results for a particular GRB to convert the foreground value into a probability which is uniformly distributed (in absence of a signal). This is then compared to a uniform distribution via either the KS test (of the median) or the binomial test. It was found that the KS test was not particularly powerful.
          Question
        : under what circumstances is the binomial test for, say, the 25% loudest events more powerful than a test of the loudest event?
      • Request for simulations showing regions in which various tests work well. Would like to see a plot like the one used in presentations to illustrate binomial test but with a number of weak signals injected to show how the distribution differs from the expected straight line somewhere in the middle (but not at the ends).
  2. SGR: no time this week. Luca is preparing the intermediate results for dissemination.
$Id: minutes-2006-05-08.html,v 1.2 2006/05/12 02:10:25 jolien Exp $