90 likes | 218 Views
Alan Weinstein (LIGO Laboratory / Caltech) For the LSC Internal review committee: Duncan Brown, Laura Cadonati, Vicky Kalogera, Bill Kells, Alan Weinstein, John Whelan LIGO LSC Meeting November 4, 2006 LIGO- G060561-00-D. Status of inspiral search reviews. What’s on our plate.
E N D
Alan Weinstein (LIGO Laboratory / Caltech) For the LSC Internal review committee: Duncan Brown, Laura Cadonati, Vicky Kalogera, Bill Kells, Alan Weinstein, John Whelan LIGO LSC Meeting November 4, 2006 LIGO- G060561-00-D Status of inspiral search reviews LIGO-G060561-00-D
What’s on our plate • Entire focus for last 9 months has been on finalizing review of six core analyses: {S3,S4)×{MACHO,BNS,BBH},to get it OUT and move on to: • S3 SBBH • and, effect of spin on non-spinning searches • S5 BNS (box opened for S5a; paper for S5-year-one) • S5 BBH (BCV PPN templates and chisq cuts, Mtot < 35) • Coherent follow-up, detection follow-ups • Parameter estimation: Time-domain-templates, MCMC • S4 (and S5) Ringdown search • GRB / Inspiral coincidence • Inspiral-merger-ringdown • We welcome more participation in review process, especially for analyses that some reviewers are too close to. LIGO-G060561-00-D
Review process • Weekly review telecons (~2 hrs) for the last year, focused on S3/S4. • Since the August LSC meeting: • “end of pipeline” cuts (“Cat2/3 vetoes” and H1/H2 consistency) • time slides as a reliable estimate of background • new technique for evaluating upper limits (Bayesian),folding in efficiency vs effective distance (with different effective distance at LHO / LLO) with astrophysical population as function of effective distance for ~fixed mass (BNS, 1.35x1.35) and as a function of mass (all 6) • follow-ups of detection candidates getting more detailed and automated • Focus on end-of-analysis numbers/results, sanity checks, etc. • Can’t check everything! But the review process forces the principal analysts to check their results, find their own bugs, etc. • Wish we had more time, but we’ve gotten comfortable with all six analyses, procedures, results… • One exception: there have been bugs discovered (by group members pursuing the analyses) associated with the “end of pipeline” cuts (“Cat2/3 vetoes” and H1/H2 consistency). In progress – numbers will likely change! LIGO-G060561-00-D
The six core analyses: {S3,S4)×{MACHO,BNS,BBH} • Common Condor DAG pipeline, common analysis procedures and codes – review essentially complete • focus code review on differences / unique aspects of each analysis– review essentially complete • Core elements of searches – review essentially complete • Detection candidate follow ups – review essentially complete • upper limit calculation (Bayesian), astrophysical population, stat errors, presentation of results – review nearly complete • end-to-end final review of all 6 analyses, key results and plots – not really done; became a paper review • omnibus Results paper – several rounds of reviewer’s comments • Methods papers(pipeline, tuning, upper limit, galaxy population) - review of papers not begun; technical papers, not sky statements LIGO-G060561-00-D
Elements of the Search that have been reviewed • Observation time (zero-lag, time lag) • Data quality flags, vetoes (Cat1/Cat2/Cat3 vetoes, r2 veto) • Template bank generation, bank size vs time • Inspiral horizon distance (AKA range, = sqrt(5)*sensemon range) • single-IFO parameter estimation; coincidence requirements • H1/H2 amplitude consistency cut, application to H1L1, H2L1 • Detection statistic: combined “effective SNR” • time slides for accidental background estimation; effective search time (in presence of vetoes) • injections, missed injections, efficiency vs effective chirp distance • Cuts reduce observation time OR efficiency • follow-up of loudest events (nearby triggers, Qscans, elog, …) • Upper limits, systematic errors LIGO-G060561-00-D
Some remaining issues / questions • Amplitude consistency test: Tricky! Several bugs found. Correctly applied for BBH? • Calculation of number-of-L10’s: 2D integral – done right? Stat error correctly propagated? • Systematic errors – correctly calculated? Waveform uncertainty for BBH? • Expect these issues to be resolved promptly…. LIGO-G060561-00-D
S3/S4 joint results paper • {S3,S4)×{MACHO,BNS,BBH} • Fairly uniform procedures used to describe all 6 searches • No plausible candidates from either S3 or S4 • Main results only for S4 (loudest events, upper limits) • A couple of iterations on paper draft with reviewers over the last 2 months • Stylistic choice: try to make this results paper short (hope was 4-5 pages, instead ~10 pages), with minimal detail on analysis procedures; detail in “companion” technical papers • Hope was 4-5 pages, instead ~10 pages. Should it be shorter? or longer? • Contrast with S4 all-sky burst paper, which makes a different choice: attempts to be self contained, while still keeping it only 25 pages (CQG format). • Anyone wanting analysis details must dig up companion papers (which may be in different journal(?)), and which are written to be self-contained, so not 100% relevant to this results paper! And, they are all in rough draft form at best. • But this “short” results paper seems to give sufficient detail to be self-contained, give results that are of interest to people only interested in the search results relevant for astrophysics • The current situation (“long” short paper) seems to be satisfactory; a matter of style. LIGO-G060561-00-D
Technical papers backing up S3/S4 joint results paper • Galaxies: Description of nearby galaxy catalog used to estimate source population (blue light luminosity as proxy for star formation rate, in contrast to some measure of mass) • Pipeline: Description of analysis pipeline used for all 6 searches (and future searches (?)) • Tuning: Description of procedures used to tune all cuts (specific to these 6 searches) • Stat&Syst: Bayesian upper limit via “loudest event statistic”, and detailed discussion of systematic errors and how they are propagated to the UL. LIGO-G060561-00-D
S3/S4 joint results paper • Reviewers have now reached a level of comfort with the analyses and with the paper. • Many small and not-so-small changes in the last few days to the paper; the paper you’ve read on the plane is already kinda old. • Bug found in the last few days, incorrect application of H1H2 amplitude consistency requirement on S4 H1L1 and H2L1 data times. Numbers will change. Probably small changes, but … Might even be a candidate! • Strong desire (shared by both analysis group and reviewers) to complete the review and get the paper out … by GWDAW! • Strongly recommend mature drafts of backup technical papers by then. • So, modulo fallout from the latest bug, addressing some remaining small issues/questions, and comments from the LSC, the reviewers are close to approving this paper. LIGO-G060561-00-D