Burst Figure of Merit

1
Burst Figure of Merit

• Julien Sylvestre
• LSC Meeting, March 2004

2
Requirements

• Real time
• lag data acquisition by less then 5 minutes
• One number
• Orthogonal to other FOMs
• Accurate
• Correlates well with our efficiency and background

3
Proposals

• Floating SNR cut
• End-to-end analysis

4
Floating SNR cut

• Process data segments with an ETG every T seconds
• For each segment, apply a SNR cut so that only N
  events survive
• FOM SNR cut
• Large glitches affect the FOM
• Small glitches dont affect the FOM much
• For a given signal model, and a calibrated
  spectrum, get hrss, range, etc.
• Ran for a month during S3

5
End-to-End analysis

• Inject a (few) waveform(s), process with ETG
• Do coincidence between all IFOs (optional)
• Measure background
• FOM sqrt(background) / efficiency
• Requires real-time calibration information
• Model dependent short/long bursts, frequency
  range, etc.
• Real-time inter-site coincidences demonstrated
  during S2

6
Orthogonality
L1, S2 playground 130-400 Hz 300s segment 100
injections per amplitude
7
Correlation between burst rate and inspiral FOM
8
Correlations
9
Conclusion

• Reasonably easy to define and implement a burst
  FOM
• Key question 1 what new information is this FOM
  generating?
• Key question 2 how good is this FOM at
  predicting the quality of the burst analysis?