An interesting experimental fact unearthed in binary compact object searches in
Refs. [<a href=''http://arxiv.org/abs/gr-qc/0505042''>Abbott:2005pf</a>,
<a href=''http://arxiv.org/abs/gr-qc/0505041''>Abbott:2005pe</a>]
is that the signal-to-noise ratios (SNRs)
of individual detectors combine in a skewed manner to form
the coherent network detection statistic. The reason for this was anticipated
by me a few years ago <a href=''http://gravity.physics.wsu.edu/boseRobust.pdf''>Bose:2002by</a>,
when I obtained the (locally optimal) network detection statistic after
maximization, and showed how the statistic is affected
by the different characters of the non-Gaussianity in the
noise of the participating detectors in a network search. My work on the inspiral
search obtained the expression for the was motivated, in turn, by Ref.
<a href=''http://arxiv.org/abs/gr-qc/9901075''>Creighton:1999qw</a>,
which analyzed
a similar effect for burst sources. The relevance of this work will become
increasingly important as multi-detector searches,
including non-LSC detectors, are conducted in the next few months (as
described in the preceding section). I am now
working with Bernard Schutz on constructing signal-based vetoes that use the
characteristics of coincident triggers in multiple detectors and the detectors'
beam-patterns to estimate the likelihood that these triggers arise from a single
astrophysical source. This likelihood can be used to discern the signals
from the background, especially, when the exact nature of the
detectors' noise is uncertain.
Robust statistics are also useful in searching for signals in LISA, which has to battle an embarrassment of riches stemming from source-confusion noise. Even if LISA's instrumental noise were Gaussian, the source-confusion noise may be anything but Gaussian. Our work on this front will be submitted for publication shortly.