Review Committee Meeting Thursday 1 December 2005 11:00 PST / 14:00 EST
Minutes: Thursday 1 December 2005 11:00 PST / 14:00 EST
MinutesWe went through the LIGO-GEO presentation focusing on the "results" slides.
- Slide 5: Perhaps "union" of times should also be given.
- Slide 7: 1s glitches in GEO: are there an excess of background events
in the subset of slides that are integer number of seconds slid?
See plot of rate vs. shift or raw data. Nothing obvious but only a cursory check.
However: Only 3% of background triggers are removed if 3ms around 1s boundaries are removed.
- Slide 8: No vetoes on these triggers.
- Slide 9: Is this too much on CorrPower? Probably not. Should
there be something about WaveBurst? Probably.
- Slide 10: Efficiencies seem only 30% to 40% larger than LIGO-only
at the same false rate. Is this a reasonable comparison (higher low-freq
cutoff so "same false rate" might not be meaningful)?
- Slides 11 and 12: Trying to understand efficiency curves. Is sensitivity
determined by worst instrument?
- Coincidence analysis in waveburst is non-trivial. Possible to have a weak signal in GEO but with high overall GC if it is strong in LIGO. But there needs to be something in GEO. Chances of this randomly happening are a few percent, but more if a long event in LIGO.
- Q: If GEO isn't being used to push down false rate to infinitesimal values, what good is it? What is the purpose of this analysis? A: Analysis goal is ultimately to do coherent analysis and to implement the null stream.
- Q: Is GEO doing nothing more than cutting false alarm rate as a random number generator could? A: It seems not... efficiency curves still go to unity; in fact, they are more "vertical" than in the LIGO-only analysis! Perhaps this is because there is better sky coverage?
- Because the nature of the 4-fold "coincidence" in waveburst is not obvious, recommend some description of the waveburst algorithm.
- Summary: we are comfortable with the material in these slides for
presentation at GWDAW. We will look over the final slides.
We went through the LIGO-only presentation focusing on the "results" slides.
- Slide 4: Igor updated us on the results of his investigation into the
effect of using the V03 calibration on the amplitude cut. See:
- There seems to be about a 1% effect on hrss50 due to getting the hrss of H2 (or H1) wrong by 10%. This is less than feared.
- For the paper (not for GWDAW), MDCs with V04 calibration will be generated and then analyzed with the V03 calibration. This will capture systematic effects.
- Note: the plot on the right of the slide has the amplitude cut applied. It needs to be regenerated from triggers before amplitude cut (since the goal is to see that few triggers are lost!). Beware that the triggers not associated with injections should be removed.
- Slide 7: Not clear that these are background. (Later, red dots are
foreground.) Should say these are background results.
- Slide 14: Curve has been removed ... should it be? Also, see clearly
the effect of the poor fit model in rightmost curve. This seems needless.
- Slide 15: Should there be some systematic factor applied to hrss50
values? We don't think this is needed: explaination that a preliminary
calibration is used seems sufficient.
- Slide 16: What about higher-frequency efficiencies? The hrss50 points
are about a factor of 10 above H2 noise curve. This is about the same
ratio as seen in previous analyses.
- Note: long duration injection efficiency curves are not being shown.
- Summary: we are comfortable with the material in this presentation. A few minor issues should be addressed. We will look over final slides.