-
Notifications
You must be signed in to change notification settings - Fork 357
PyCBC Live O3 Review
The major difference from the O2 analysis is that triggers are now generated from all detectors. This works by first generating double coincidences from all possible detector pairs, then picking the most significant double, and finally using the SNR time series in the remaining detectors to derive a p-value, which is then used to modify the FAR of the original double coincidence.
(See PyCBC-Live-O3-development for more discussion in context of the overall development plan.)
Procedure: let the pipeline run on O2 replay data and monitor the duty factor, lag and memory usage, making sure all combinations of detectors are explored.
Contact: Tito
Outcome:
- Daily lag monitor for replay analysis and production analysis.
- Memory usage appears stable.
Signoff:
Procedure: plot the rate of triggers vs their FAR after running on replay data.
Contact: Tito
Outcome: preliminary plot showing agreement
Signoff:
Procedure: using the O2 replay analysis, wait for triggers to be uploaded (with all possible ifo combinations) and check that BAYESTAR completes successfully.
Contact:
Outcome:
Signoff:
Reference could be the O2 PyCBC Live configuration, or O3 offline search.
Procedure: TBD. Reuse the data/scripts for the O2 review?
Contact: Bhooshan
Optional?
Procedure: repeat the analysis done on step 4 with a different combination of detectors.
Contact:
- Pregated/ungated strain
- State/DQ channels and flags
- SNR threshold, NewSNR threshold, GraceDB upload threshold
- Choice of ranking statistic
Signoff:
Procedure: banksim
Contact: Soumen?
Signoff: