LSC-Virgo Burst Analysis Working Group
Special Review Meeting Tuesday December 1 2009 08:00 Pacific / 11:00 Eastern
Minutes: Tuesday December 1 2009 08:00 Pacific / 11:00 Eastern
Agenda and Contact Info
- Omega on-line analysis
On EVO use "bersten" to get in.
- Present: Owen, Kanner, Klimenko, Searle, Raffai, Flaminio, Rollins, Hughey
- Searle: Old whitening was simplistic, contaminated by loud injections so SNR saturated at 200-300 depending mainly on frequency. X-pipeline used it too; Sutton and Edwards fixed using findchirp method "mean median power" or similar.
- Searle: Old algorithm didn't truly matched filter SGs either due to whitening (denominator of filtering integrand); now it does.
- Searle: Now clustering based on Chatterji's paper.
- Rollins: Much re-architecture of pipeline to improve efficiency, verified no effect by re-running old results. Klimenko/Owen/Flaminio: Did you guys check Antony's changes on JW1 etc? Not really. Not catastrophic unless it shows in live test, move on.
- Searle: Parameters are sine and cosine amplitudes of + and x. Bayesian C code gives physical parameters, converts those to response parameters. Assume uniform over sky and arrival time, 4 amplitudes are gaussian distributed. Doesn't quite use proper distribution for circularly polarized waveform but is close, much simpler to integrate.
- Searle: Big prior assumption is SG waveforms! Could generalize later.
- Klimenko: Residuals don't work well as veto (when comparing to noise).
- Owen: Overcomplete basis could lead to funny statistics? Searle: Thinks OK, let's revisit after seeing coherence etc.
- Disagreement over whether Omega is missing many injections or passing more non-injection triggers.
- Kanner: cWB was based on all S6 background, Omega more recent. So cWB was trigger far less than desired rate. cWB has changed to more recent-data background estimate last night, should see in LUMIN from now on.
- Rollins: Nov 27 was bad, restarted analysis and threshold was too low.
- Flaminio: Where do pipeline figures of merit come from?
- Owen: How does sleepy human compare 98% Omega to cwB rho? Comparable backgrounds are needed.
- Searle: Much new linear algebra, PRC tells you it's correct. Coherent followup (Raffai) uses detector time series to construct skymaps, really shows no silly math error. Adaptive mesh on highest probability pixels.
- Owen: Given multiple fringes, does LUMIN send multiple pointings?
- Kanner: No ROTSE, yes QUEST.
- Owen: Off topic, but: Why did cWB look better on WNB than SGCs? Searle: Klimenko is gone, but think it's because they accumulate info across multiple basis functions?
- Flaminio: What SNR are triggers you're sending 1/day? Various: Not simple answer, we use other statistics.
- Flaminio: Recently LUMIN shows both pipelines seeing injections, good. But background estimates are rather different, what gives? Rollins: Not sure, need to look into it. They background on different stretches of data.
$Id: minutes-2009-12-01.html,v 1.3 2009/12/09 22:40:55 owen Exp $