© Warren G. Anderson, The University of Texas at Brownsville, 2003.
Looking for Excesses of Power
- method described by Anderson, Brady, Creighton and Flanagan (IJMPD9, 2000; PRD61, 2001).
- search time-frequency map for rectangles with too "much power", such as black rectangle at left.
- central limit theorem implies that random fluctuations (noise) from many different sources, whatever their individual distributions, will form a normally distributed population.
- power is sum of squares of normal deviates, which is distributed as a χ2 variable.
- set power threshold so that false alarm rate is low enough, i.e. so that χ2 probability is acceptably low.
- use coincidence criterion in several detectors to further lower false alarm rate.
- can prove that this is an "optimal" detection strategy when only duration and frequency band are known.
- currently used to look for black hole coalescence.
- major strength: almost no information about signal needed.
- major weakness: every instrumental glitch registers!