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

Minutes of 2008-Feb-07 S5 QPipeline Review Teleconference


Shourov Chatterji, Jonah Kanner, Isabel Leonor, Dave Reitze

Minutes by Dave Reitze.



S5 H1H2 1 day playground

The investigation of trigger distributions has been extended to include estimates of the significance of the zero lag triggers relative to the non-zero lag background. Tables of the loudest events are now available, including QEvent and QScan analysis. Scatter plots of the coherent and correlated energy of injections, non-zero lag trigger, and zero lag triggers are also under study, and may lead to an improved definition of significance. All of the reuslts are posted here.

S4 H1H2 analysis

The S4 H1H2 anlaysis has been similarly extended. All of the results are posted here. A comparison of the resulting efficiencies and upper limits with the existing triple coincident WaveBurst/CorrPower results indicates comparable performance for the SG20_V4_S4 injection set of simulated sinusoidal Gaussian signals with Qs of 3 and 9. Simulations have not yet been performed for the other simulation sets.

S5 H1H2 1 year non-zero lag trigger production

Trigger production is complete for the 1 second time lag analysis. The resulting distributions are posted here.

Triple coincident analysis

L1 triggers are available for the S5 1 day playground, including simulated signal injections. As a result, work on the triple coincident playground analysis can start immediately. L1 triggers are also available for all of S4, but not for simulated signal injections.


Shourov: code is now frozen and will tag it in CVS; it hasn't been tweaked in a
while.  Code review team will look at code after this in CVS.

Supposed to be looking at S5 1 yr zero-lag data, but not done running yet.

Jonah: harking back to X-pipeline call, difference between zero-lag and non-zero
lag significance is worth getting into more deeply.

Shourov: yes, will go into this later.

Start with S5 H1H2 1 day playground and extension of trigger analysis.  Posted
analysis segments and livetimes.

Shourov walks us through zero and non-zero lag trigger distribution for
different applications of vetoes.

Has made tables of 10 loudest events for each case; Q-scans done on events.
Clicking on rank displays summary of event -- data quality summary included.

Next Shourov walks through scatter plots to compare correlated vs coherent SNR
-- want to see if one is better than the other?  Also considers injections with
the criterion that H2 is half as sensitive to H1.

A quick re-walk through of detection efficiencies for Jonah.

Isabel -- should plot number of events in addition to false alarm rate so that
we can do the statistics ourselves.

Jonah -- should we be troubled by the event that pops out?  Maybe we shouldn't
worry because this is a loudest event statistical approach to upper
limits. There was some discussion of this last week, in particular the role of
data quality flags and follow ups.  Shourov points out that this event doesn't
show up on L1.  Some discussion took place at the Burst group (eg look at 10th
loudest event statistic instead of loudest event)

A discussion of S4 results -- notice one event really pops out.  This is a
'power-mag' glitch, a power line transient.  Flag was not in place for S4, but
hopefully will be for S5.

Neglecting that, the H1H2 event upper limit is compared to WaveBurst/Corrpower
H1H2L1 result -- results are comfortingly similar.

Dave -- should fold L1 into S4?  Shourov -- there are computing power issues.
Maybe worth doing after S5 analysis if we want a sanity check.


While we were talking, the S5 H1H2 1 year non-zero-lag run finished, so we had a
quick look at 1 yr of S5 non-zero-lag data with 1 s time shift -- look at H1H2
correlated vs coherent SNR.

Jonah -- interesting that the distributions are similar to 1 day playground.
But, here we have only one time shift and in the playground set we have 200 time
shifts. Also, where correlated SNR is '0' on plot, it's really negative?
Shourov: Yes.  Also note that anything below an SNR of 4 will be thrown out by

Finally, everything is ready for full triple coincidence S5 playground data.
That work should go fast.  For S4, not ready yet, but worth doing.

Next week: should focus on triple coincidence playground analysis

Dave: possible to have materials ready 1 day in advance for review committee to
digest?  Shourov -- sure, should be no problem.  Will send around a pointer.

Action Items

$Id: minutes_20080207.html,v 1.2 2008/02/07 22:50:46 shourov Exp $