coherent WaveBurst Review Meeting 02 Apr 2009 09:00 Pacific / 12:00 Eastern
Agenda: Thursday 02 Apr 12:00 Eastern
- (New) Review Tasks and Action Items page: [web].
- Search results pages (linked from Igor's 2cd year page) [ web ]
Dial-in number: 1 800 704 9896 International dial-in: +1 404 920 6472 Conference code: 5374 2349 #
AttendancePatrick (minutes), Marco, Francesco, Igor, Sergey, Virginia, Stas, Romain.
Search results pages
- Based from Igor's 2cd year page. Page contains details of analysis of each network, including links to Sergei's results pages for background and zero lag.
- IY: Sergei's pages were generated using the original DQ segments. The livetime has to be corrected for the final segment lists.
- IY: Both the background and zero-lag pages contain efficiency curves. They should be the same on both.
- Focussed initially on L1H1H2 network.
- Summary statistics section: Zero-lag triggers available under the "zero lag events" link. Table in this section shows only time-lag triggers above final threshold.
- Zero-lag page: Mis-flagged inspiral injection visible as rho=5.5 trigger. Equinox event is rho=6.6 trigger.
- Simulation section contains link to file describing processing of triggers to produce -1/0/1 lists of injections for use in making combined upper limits.
Readme file applying final segments list to injections: README. Steps:
- split.py splits injections by waveform type and strain.
- split1 does same for all injections regardless of whether they are in final seg list or not.
- Matching done under condor to classify as detected (1), missed (0), not considered (-1)
- eff.py makes efficiency data files
- mdc1.C plots efficiency curves and does fitting
- hrssN1.py computes hrss 50% and 90% values.
- patrick.py (!) generates the files needed for combining results from several pipelines into a single UL. PJS: This is probably the most important for review, since it goes into result in paper.
- page.py generates webpage
- Next looked at H1H2 pages.
- IY: H1H2 is a special case in that some zero-lag data was used as playground for tuning (to check for correlated background). Relatively high threshold chosen to be conservative.
- Section 1 (production): rate vs. lag plot shows excess at first non-zero lag (3.125 sec). IY: Threshold for these triggers is very low, rho>3.0. Excess associated with low-rho triggers; disappears for higher threshold.
- Analysis pages carry links to text files specifying rho thresholds as a function of time and frequency. Links have names like "T[4.1,6.0,5.0]".
- Action item" IY to post link to zero-lag playground analysis page for H1H2.
RG: Seem to be too many -1s in the WNB -1/0/1 injection file, considering the >50% duty cycle, even ignoring 4X times.
Action item: IY to double-check this file.