LSC-Virgo Burst Analysis Working Group
X-Pipeline Review Meeting Monday 10 March 2008 09:00 Pacific / 12:00 Eastern
Minutes: Monday 10 March 2008 09:00 Pacific / 12:00 Eastern
- We finished the technical "Notes on Coherent Analysis Techniques in GWB Searches" [ PDF ].
- E_NULL, E_TOT, E_SL. Suppose just have H1 and L1 (two detector,
not aligned case). What happens then?
A. E_TOT is the same as E_SL. There is no NULL stream.
- LIGO network: "trivial" to create a coherent combination of H1
and H2, so have effective two detector network with LHO and LLO.
Yes, then NULL is h_H1-h_H2, and E_TOT is the same as E_SL with
the two effective detectors.
- So, for LIGO network, E_SL and E_NULL are not really sky-position
dependent (apart from time shift issues)? Why problem of searching
over sky positions?
Yes, but X-Pipeline doesn't use E_SL ... it uses E_+ discussed
later. Also, really want this to be a method that uses
larger networks (though the NULL stream is of significant
benefit even within the LIGO-only network).
- Patrick will send maps of |F+| and |Fx| for the LIGO network to
illustrate the dominant polarization frame.
- Q. When there are lots of instruments with similar sensitivity
and various orientations, then presumably |F+|~|Fx| and so the
dominant polarization frame is only weakly dominated.
- Discussion on dominant polarization frame.
- Q. Why choose E_+ rather than E_SL? Isn't E_SL, the max
likelihood method, the "optimal" method?
A. Actually, the Bayesian optimal would involve the integral over
possible sky positions.
But these are weak integrals, and the exponential of the likelihood
is strongly peaked, so shouldn't the max likelihood point dominate?
Regions where the X-polarization matters are isolated near
detector nulls, where the sensitivity is poor and these
regions are not favored in integral. There is an AMALDI
proceedings discussing this.
Patrick will send AMALDI proceedings.
- Criticism: even if E_+ is empirically better than E_SL now,
will it still be so when other sensitive detectors are added to
the network? At what point does one "flip-the-switch" and go from
E_+ to E_SL? Should one use soft-constraint likelihood which
smoothly dials from one to the other?
Soft constraint had problems in all-sky search, but might work
well for GRB search. (Problem was that the regions near
the detector nulls always dominated statistic -- not a
problem for fixed sky location problems.)
$Id: minutes-2008-03-10.html,v 1.1 2008/03/10 20:00:41 jolien Exp $