S4 LIGO-GEO Burst Review Committee Special Meeting 29 April 2008 10:00 Eastern
Agenda and minutes of the special LIGO-GEO S4 Bursts Review telecon, held Tuesday 29 April 2008.
Agenda and Contact Info
- Discussion of problems with efficiencies in the iWB-CP pipeline:
- Inconsistency of Fig 5 with Table 2.
- Igor's estimated iWB-CP efficiencies do not go to 100% at large amplitude.
InterCall telecon service: USA (toll-free): 1-866-616-1738 UK (toll-free): 0800 073 8914 Italy (toll-free): 800-906-494 Germany (toll-free): 0800-1014-907 participant code: 251 288 9495#
AttendanceLaura Cadonati, Siong Heng, Patrick Sutton, Keith Riles, Sergei Klimenko (minutes), Igor Yakushin.
Statement of the problemsProblem #1: a discrepancy between Figure.5 and Table.2 in the LG draft. For example, the figure 5 shows hrss@50% = 6.2e-21 for SG849Q9 and the table.2 states hrss@50% = 7.0e-21. This is one problem and it is not related to what Igor is doing. As I understand correctly, both the Fig.5 and the table.2 for incoherent analysis were produced by Siong and you. So you, guys, can check what is wrong: the plot, the table or both. My guess would be that the plot shows efficiency for non-resampled injections and the table shows efficiencies for resampled injections.
Problem #2: a discrepancy between the Figure.5 and the tables produced by Igor. By the reviwers request Igor tried to reproduce the efficiency curves for the incoherent analysis and he obtained a different result. The nature of the discrepancy is that from the figure.5 it follows that loud injections (hrss~3e-19) are detected at very close to 100% efficiency and Igor's tables http://ldas-jobs.ligo.caltech.edu/~igor/test_032808a/ show that they are detected at 96-97%.
Status and Action ItemsProblem #1 is resolved. The problem is that Fig.5 was generated using an separate fit to the sampled efficiencies, separate from that made by Laura for the numbers in Table 2. Patrick has made a short matlab function that computes the efficiency curves at user-specified hrss given Laura's sigmoid fit. He verified that using the fit parameters provided by Laura, the 50% and 90% points read from the graph are a close match to those listed in Laura's parameters file and Table 2. Siong is going to do the following:
- Produce a new figure 5 using Laura's fit parameters.
- Check the new figure 5 for consistency with table 2 for the incoherent analysis.
- Check matlab fitting procedure for the coherent analysis.
- for incoherent pipeline the burst MDC log (with injections at 2.5e-21 strain) was resampled by Brennan. For example, given an injection at certain gps time it is either excluded from analysis for all injection strains or used for all strains.
- for coherent pipeline the coherent mdc log (which includes multiple injections at different strains) was resampled. For example, given an injection at certain gps time, some injection strains are excluded and some not.
- Igor will check if there is a discrepancy (and how big) between final injection lists for coherent and incoherent pipelines. It will check if the same data is used for estimation of the sensitivity. Were the same DQ flags used?
- Siong will check what DQ flags were used in the incoherent analysis and whether the bad 10 days period was included or not.
- Igor will regenerate cWB efficiency curves for the resampling procedure used for the incoherent analysis and S4 paper (i.e., using Laura's injection lists, downsampled to the cWB injection times). Regenerate Fig 7, Table 2.
- Igor will generate also cross-check tables for the incoherent pipeline from Laura's trigger lists.