X-Pipeline Review Meeting Monday 08 Sept 2008 09:00 Pacific / 12:00 Eastern
Agenda: Monday 08 Sept 2008 09:00 Pacific / 12:00 Eastern
Recap of xcondition.m review.
F2F presentation: Status of analysis and proposal for completion [web]
Code walk-through: m2html documentation for revision 2301 [ web ]
Attendance: Patrick, Keith, Jolien, Antony
- Conditioning code -- seems a bit dodgy.
- Antony was skeptical of LPEF code. Suggest rewriting the
- New code written, unable to test due to lack of matlab.
Test of conditioning code to be resolved. Periodogram based.
Don't use LPEF coefficients. Median power across time bins.
Gives power spectrum. Compare with FINDCHIRP paper:
- F2F presentation: Status of analysis and proposal for completion [web]
- 81 closed-box results, remaining 25 for LSC meeting.
- Improvements - injections into off source; Calibration
fixes; Sigmoid fitting; PSD estimation; Tweaks;
- Takes 3 months to get through the GRBs! How to deal
with these changes? 1 min per run. 12 waveforms x
25 ampls x 500 times = 1 day per GRB.
- Proposal: rerun on/off source processing. 99% of time
is in injections. Simulations on only 10 of the GRBs.
- Jolien doesn't like this -- would rather reduce number
of simulations by a factor of 10 (increase random error)
rather than trying to guestimate systematic error.
- Is it worth it when there is a fixed calibration
- Use of ULs? What is the scientific payoff of these ULs?
Perhaps just for a subset.
- Suggestion: just cut down on the number of simulations
done (waveform types, number of trials per type). Increase
random error but no systematic error (due to code change).
- Note: cut thresholds are based on simulations too.
Don't want to increase statistical error on estimating these
- Plan is to present only two waveform ULs anyway.
Why not just do these rather than the full 12 waveforms
for publibcation purposes. Reduces number of simulation
runs required by factor of 6... now it is doable in a couple
$Id: minutes-2008-09-08.html,v 1.3 2008/09/08 17:11:18 jolien Exp $