Accelerating parameter estimation of gravitational waves from black hole binaries with reduced order quadratures

The inference of binary parameters from gravitational waves (GW) is one of the key science goals of the collaboration operating the advanced ground-based LIGO-Virgo detector network. We employ reduced order quadratures (ROQs) to substantially reduce the size of large inner products arising in Bayesian parameter estimation (PE) and thus enable studies of the GWs emitted by coalescences of spinning stellar mass black hole binaries approaching the full design sensitivity of these detectors. We build the first ROQs that include the inspiral, merger and ringdown parts of the GWs for a single-spin precessing phenomenological waveform model (IMRPhenomP) and for an aligned-spin effective-one-body model (SEOBNRv2). The ROQs for SEOBNRv2 use a separate reduced order model (ROM) as a proxy. The ROQs we have constructed are suitable for any power spectrum density function of ground-based GW detector noise. We find speedups in the calculation of inner products and the likelihood function of up to several hundreds, reducing to days analyses which could otherwise take up to a year to complete.