Review Committee Meeting Thursday 10 August 2006 09:00 PST / 12:00 EST
Minutes: Thursday 10 August 2006 12:00 PST / 15:00 EST
Agenda and Contact Info
- Update on SGR Search: see the following links
- Search sensitivity [
- On-source results [
AccuConference teleconferencing service:
Phone: 1-800-704-9896, participant code: 038621#
International callers ++1-404-920-6472 with same code
- Note: minutes are entered long after the telecon took place, so they
- Major concerns were (1) that the power tends to be negative and (2)
the frequentist upper limit on the GW energy can occationally be negative.
These are unrelated issues.
- Negative power: this is relative to adjascent bands. Since
noise power spectrum is convex, the average power in the two adjascent
bands will exceed the power in the on-source band. Thus, the negative
powers are expected.
- Upper limit is negative. This can happen if the on-source time
happens to be peculiarly quiet. See for example the distribution for the
92.7Hz RHESSI QPO (150s-260s) SG Q=1e6 [
How does one handle this situation? Several options:
No resolution at this time for the best way to proceed.
- The negative limit is valid, if unphysical -- it is simply a
manifestation of the queer frequentist definition of the limit.
- There are "traditional" frequentist methods for dealing with this.
See the section on Statistics from the 1997 PDG [
- Feldman-Cousins method also deals with this issue -- this has become
somewhat standard but hesitant to recommend it because we would not want
this method to become the basis for concluding a detection (in the case
of an interval bounded away from zero).
- Bayesian methods do not encounter a problem, but a prior would need
to be selected.
$Id: minutes-2006-08-10.html,v 1.3 2006/09/11 14:19:38 jolien Exp $