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

Minutes of 2008-Feb-21 S5 QPipeline Review Teleconference

This call was rescheduled for 2008-Feb-22 at 9:00 PST / 12:00 EST / 18:00 CET.

Attendance

Shourov Chatterji, Jonah Kanner, Isabel Leonor, Dave Reitze

Minutes by Isabel Leonor.

Announcements

Agenda

Status of trigger production

Trigger production is continuing for the first year of S5. Due to relatively heavy demand on the Caltech computing cluster, much of the anlaysis is being run on the UWM Nemo cluster.

RAW
  H1H2 timelag  CIT   done
  L1            CIT   done

HGA2_S5_R2
  H1H2          UWM   done.  needs postprocessing.
  L1            CIT   done.

HSG1_S5_R2
  H1H2 1.0      CIT   done
  H1H2 other    UWM   done.  needs postprocessing.
  L1            CIT   done

HSG5_S5_R2
  H1H2          UWM   done.  needs postprocessing.
  L1            UWM   done.  needs postprocessing.

HSG6_S5_R2
  H1H2          UWM?  will start when HWNB2_S5_R2 complete
  L1            UWM?  will start when HWNB2_S5_R2 complete

HWNB2_S5_R2
  H1H2          UWM   running started feb 21 @ 11:00 CET
                      40% complete on feb 22 @ 10:15 CET
  L1            CIT   running started feb 20 @ 21:15 CET
                      39% complete on feb 22 @ 10:10 CET

H1H2 time lags of +/- 1, 2, and 4 seconds have been tested.
If time allows, consider running more timelags of +/- 8 and
16 seconds, as well as non-integer second shifts and larger
shifts.

Coincidence window duration

An investigation was performed on the effect of HL coincidence window duration on detection efficiency for the S5 1 day playground. The results indicate that there is little advantage gained by extending the coincidence window beyond 10ms. At the same time the false rate scales with the coincidence window. As a result, a concidence window equal to the speed of light travel time of 10 ms is indicated.

Data quality and vetoes

A closer look at the quality of L1 data during the S5 H1H2L1 1 day playground indicates that this was not a good day for L1. This can be seen from the L1 QOnline analysis (day 1, day 2), as well as data quality flags. In particular, the loudest L1 event that was seen to be in time-lagged coincidence with an H1H2 zero lag trigger would be removed by category 2 and 3 data quality flags. The latter flag vetoes almost an hour of data towards the end of a particularly bad lock segment, and may be associated with a train according to detector logs.

It is also interesting to note that, based on an email to the burst group from Igor Yakushin, the loudest event from the 1 day playground shows up as significant event in the Coherent WaveBurst H1H2L1 time lag background study. Presumably they have held H1H2 at zero lag for this triple coincident study. As a result, there time lag triggers to some extent trace out the shape of the H1H2 zero lag data set. This seems to indicate that the playground day is not representative of the full year in the sense that it may contain the loudest H1H2 event. There is still the possibility of a louder event that did not happen to be coincident with a time slide, or a louder event in the H1H2 set when L1 was not in science mode.

Links

S5 playground
S4
S5 1st year

Minutes


* Dave:  Talked to Sergei, and WaveBurst does things differently.  No time shift
  for H1H2, but time shift with respect to L1.  Sergei says this has been
  standard analysis.

  Shourov:  Standard for past analyses.

  Jonah:  The point is that for coincident analysis, H1 and H2 are correlated.
  Using time slides between H1 and H2 can lead to artificially low statistics.

  Shourov:  Have not really looked at 3-detector analysis.  For 3-detector
  analysis, using zero-lag H1H2 is really a semi-open box.

* Shourov then started to go through the agenda items:
  http://www.lsc-group.phys.uwm.edu/bursts/review/projects/s5-qpipe/minutes/minutes_20080221.html

  Shourov:  Trigger production looks good.  Six H1H2 time lags.  Multiple scale
  factors for injections.

  Jonah:  For combining results, the different analyses might have to use the
  same injection scale factors.

  Shourov:  Can check for consistency of scale factors.

* Shourov then discussed results of study using different LHO-LLO time 
  coincidence windows:
  http://ldas-jobs.ligo.caltech.edu/~shourov/s5qreview/playground_withveto/H1H2L1/window/

  Shourov:  Results show that 10-ms coincidence window is sufficient.  Even
  increasing window to 100 ms has minor effect on efficiency.

  Isabel:  Were all the tests done using linearly polarized injections?

  Jonah:  Can you document what the cuts are for doing coincidence?

  Shourov:  DQ issues have not been addressed.  Chose cut at one background
  event where he thinks things are going to be.

  Jonah:  Again, for transparency, is it possible to write down the rules for
  when tiles are considered overlapping?  What tiles are considered 
  overlapping?

  Shourov:  These are in the technical documentation, under collocated
  analysis:
  http://ldas-jobs.ligo.caltech.edu/~qonline/s5/1.0/documentation/technical/ 
 
  For null veto, the definitions are similar; the difference is that the alpha
  factors increase as a function of null energy.

  In the coincident analysis section of the technical documentation, T=10 ms
  and F=0.  Use max function ensures that one tile overlaps the center of the
  other tile, i.e. tighter test.

  Category 2, 3 vetoes are available.  KleineWelle vetoes will soon be
  available.

  Jonah:  Have you thought about the cuts you will be making to get to the
  loudest event?

  Shourov:  Have been discussing ideas with Patrick Sutton, to make use of
  probability distributions.

  Shourov:  Are noise and injection triggers treated equally?  No.

  Shourov:  Calculation of incoherent energies giving out negative numbers.
  Have to make sure that these values are treated correctly.  This is due to
  assuming flat noise spectrum.  Correct treatment is important for null
  stream, not so important for signal stream.

  Shourov discussed difference between sparse triggers, i.e. where most
  significant trigger was kept on 1-second time scale:
  http://ldas-jobs.ligo.caltech.edu/~shourov/s5qreview/playground_withveto/H1H2L1b/

  and non-sparse triggers:
  http://ldas-jobs.ligo.caltech.edu/~shourov/s5qreview/playground_withveto/H1H2L1c/

  Shourov:  Difference in sparse, non-sparse is there for low-Q, not a problem
  for high-Q.

  Isabel:  Are the raw triggers in "c" plots also non-sparse, like the
  injections?

  Shourov:  No.  This goes back to the issue that noise and injections are not
  treated the same.

  Shourov:  Things to be done:  Vetoes; definiton of H1H2L1 significance.

  Dave:  In analysis telecon, P. Shawhan says that box could be opened by
  March meeting.

  Shourov:  That's the intent of upcoming F2F meeting, to discuss the
  timeline.

  Dave:  Are you still on track to do that?

  Shourov:  Open box after March meeting, not before.  Will discuss things in
  F2F.

Action Items

$Id: minutes_20080221.html,v 1.5 2008/03/06 07:49:46 leonor Exp $