Hierarchical Bayesian method for detecting continuous gravitational waves from an ensemble of pulsars (1807.06726v2)

Abstract: When looking for gravitational wave signals from known pulsars, targets have been treated using independent searches. Here we use a hierarchical Bayesian framework to combine observations from individual sources for two purposes: to produce a detection statistic for the whole ensemble of sources within a search, and to estimate the hyperparameters of the underlying distribution of pulsar ellipticities. Both purposes require us to assume some functional form of the ellipticity distribution, and as a proof of principle we take two toy distributions. One is an exponential distribution, defined by its mean, and the other is a half-Gaussian distribution defined by its width. We show that by incorporating a common parameterized prior ellipticity distribution we can be more efficient at detecting gravitational waves from the whole ensemble of sources than trying to combine observations with a simpler non-hierarchical method. This may allow us to detect gravitational waves from the ensemble before there is confident detection of any single source. We also apply this method using data for 92 pulsars from LIGO's sixth science run. No evidence for a signal was seen, but 90\% upper limits of $3.9\ee{-8}$ and $4.7\ee{-8}$ were set on the mean of an assumed exponential ellipticity distribution and the width of an assumed half-Gaussian ellipticity distribution, respectively.