LIGO Mock Data Challenge
In an effort to prepare for upcoming analysis of LIGO data, it is planned
to create and analyze simulated LIGO data containing both realistic detector
noise and one or more signals from astrophysical sources (known only to
one or a few physicists sworn to secrecy).
LIGO physicists developing astrophysical source searching algorithms
could then search for signals and report on their findings without knowing
the "real answer". This mock data challenge is expected to stimulate analysis
work and to serve as a valuable diagnostic on that work from which future,
real LIGO analysis will benefit.
The mock data set will include simulated data from at least two of the
three LIGO nterferometers and perhaps simulated data from other gravitational
radiation detectors worldwide. Part of the effort will include generation
of mock meta-database information similar to that planned for LIGO.
As preparation for this planned challenge, data taken in November 1994
with the Caltech 40-meter prototype interferometer (without ) will be analyzed.
Despite many drawbacks (e.g., no recycling or optical recombination
and only a handful of DAQ channels recorded), this data set is valuable
because it is real data from a LIGO-like suspended interferometer.
Its noise has many non-Gaussian components, including fast transient
glitches. As a result, the data set presents a serious challenge to analyses
that would search for astrophysical sources. It should serve as a hard-knocks
training ground for LIGO algorithms.
The initial effort will concentrate upon detector characterization (e.g.,
power spectra, transient classification / identification, correlation identification,
As the data set becomes better understood, it will be split into two or
more segments, with the segments treated as if coincident in time. This
will allow tests of coincidence algorithms and a direct measure of their
Finally, artificial astrophysical sources will be introduced into the
data set and physicists challenged to find them, as described above.
Additional data from the 40-Meter prototype may become available in early
1999, data taken in a fully recycled configuration and with many DAQ channels
recorded. This dataset could be used in the same way as the 1994 data set.
The End-To-End Model should
allow more realistic simulation of LIGO's Gaussian instrumental noise,
perhaps to be mixed with non-Gaussian (worst-case, one hopes!) noise
found in the 40-Meter data.
Eventually, using what is learned from the 40-Meter data, from the End-to-End
model and from other special-purpose models of the interferometer, a true
LIGO-like mock data challenge will be mounted.
This challenge has been initiated by the LIGO Laboratory, but will be carried
out in close collaboration with the LSC. In particular, the LSC data analysis
working groups (Astrophysical
Source Identification and Signatures, Detector
Characterization, and Detection
Confidence and Statistical Analysis) will naturally play a strong role
in the challenge. This challenge, however, is open to the gravitational
wave community at large, which will be given free access to the November
1994 40-Meter data.