LIGO Seminar

Current studies into the accuracy of inference on gravitational-wave signals rely on either computationally inexpensive, but often inaccurate, predictive techniques, or on complex stochastic Bayesianmethods able to make robust inference based on noisy signals but at a significant computational cost. Here I will argue that it is possible to have the best of both worlds, a fully Bayesian approach able to produce posterior probability distribution functions at the same accuracy as the stochastic methods, but carried out on a simple grid at a low computational cost, comparable to the inexpensive predictive techniques. One area where efficient parameter estimation will be crucial is for classification of CBC sources into subpopulations (BNS, NSBH and BBH). Considering a astrophysically motivated distribution of merging, and detectable, binaries produced using population synthesis methods along with measurement accuracy estimates from the main gravitational-wave parameter-estimation pipeline, I find that the BNS and NSBH subpopulations will always be distinguishable. This will not necessarily hold at the NSBH-BBH boundary, but with a few tens of detections the three subpopulations can accurately be identified together with measurements of their respective rates.

We plan to broadcast this talk using SeeVogh.