Marginalisation

Gravitational Wave Phase Marginalisation

In the bilby example, the time-domain waveform from this paper is coded up (using aligned-spins) and used as the waveform model, and posteriors are produced with 4 free parameters. However, the phase is marginalised out of the likelihood in the ‘pre-sampling stage’, giving the sampler one less parameter to search over and thus reducing the cost of computing the phase-marginalised posterior by a factor of a few (compared to marginalising post-sampling). dolfen has an inbuilt (log-)likelihood function that is designed to receive individually the plus and cross polarisations of the the waveform from the GWmodel_margphase callable, and is activated when a GWmodel_margphase callable is passed as a parameter to dolfen. This allows for a quick and easy pre-sampling numerical marginalisation/integration over the phase. To use a different model with a (pre-sampler) numerical marginalisation, see marginalisations for arbitrary models.

Marginalisations For Arbitrary Models

Note

This section has not been completed yet