LSC-Virgo Burst Analysis Working Group


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


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


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


Main review page

Review Committee Meeting Monday 19 June 2006 09:00 PST / 12:00 EST

Minutes: Monday 19 June 2006 09:00 PST / 12:00 EST

Agenda and Contact Info


  1. Continue discussion of vetoes used in the SGR1806-20 search [ HTML ] and look at preliminary results from relative band excess [ HTML ]
  2. Continue discussion of draft paper for the S2-S3-S4 GRB-GWB search [ PDF ]

Contact Info

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


  1. Discussion of replacement for band correlation veto.
    • APS results used two vetoes, an 8-sigma veto on power and a band-correlation veto.
    • Proposal to do-away with the band-correlation veto: compute power in on-source band and two adjacent bands. Compute excess power of on-source band relative to the average of the power in the two adjacent bands.
    • Performance improves by about 20% from 1e-22 to 0.8e-22 largely because the veto had previously rejected approximately 20-22% of the data.
    • Brian: worried about up-conversion -- what if there is up-converted noise in one band but not another? Possibility, but this would still be correctly accounted for by off-source background estimates. Issue might affect sensitivity of search but not validity of result.
    • Major purpose of this method is to account for the "breathing" which would affect the on-source band and ajacent bands similarly.
    • Will the group try to optimize the other veto too? Yes.
    • In reality, analysis method divides the long (~50s) segment into short (~0.25s) segments, computes the power in each of these, and adds them up. Perhaps instead of taking (on-source power) minus (average of ajacent band power) you could do something like the following:
      • For each 0.25 seconds in the 50 second segment
        • Compute power in on-source band
        • Compute average power in off-source band
        • Compute ratio of (on-source power) / (average of ajacent band power SQARED)
        • Add this value to the cumulative "effective on-source power"
      With this method, periods of high noise power will be naturally suppressed relative to periods of low noise power. (You can probably show that this is the optimal method if the signal amplitude is assumed to be constant.)
    • Perhaps with this method, the first veto can be eliminated too? Szabi prefers to keep the first method because it is a clean rejection of very short intervals that are no good.
    • Luca will try implementing this method too to see if it is effective.
    • Schedule: Sharmilla is completing code review, but there is now some new code. She will have a document by next week which she will give to Shourov who will look things over with a fresh set of eyes.
    • Goal is to have a draft by the LSC meeting (sooner actually).
  2. Continued discussion of GRB search.
    • Jolien's first comments on draft are now available in electronic form PDF ]
    • Isabel is running jobs on the circular polarization studies.
    • Issue with draft: there are 3 different statistical tests of the null-hypothesis. This needs to be justified. Either justify claim that the methods are "complementary" or suggest that some of the methods are exploratory (and then comment on our conclusions about them). Furthermore, the ad-hoc choice of using the loudest 25% of events in the binomial test needs to be justified at some level (e.g., why this is better than just looking at the loudest, or at all the events). Need a simple population model to show complementarity of methods.
    • Calibration: S3 and S4 used most recent calibration. Uncertainties for S2 should be done the same as the untriggered search. Uncertainties in later calibration is less. S5 now uses DARM_ERR.
    • Schedule: not sure how much more effort to put in the S2-S3-S4 results with S5 results available. First epoch of S5 to be released is up to 3 April 2006 and there are ~50 GRBs in this epoch. Perhaps scrap some of the analyses for this paper and put them in the S5 paper? But studies would need to be in either the S2-S3-S4 paper or the S5 paper -- not clear if there is any net delay (of S5 results) by doing them in the earlier paper.
$Id: minutes-2006-06-19.html,v 1.2 2006/06/20 14:13:53 jolien Exp $