Marginalizing over noise properties in parameter estimation

Other Versions:

Abstract:: The traditional gravitational wave parameter estimation process relies on sequential estimation of noise properties and binary parameters, which assumes the noise variance is perfectly known. Using new capabilities of the BayesWave algorithm and recent developments in noise uncertainty modeling, we simultaneously estimate the noise and compact binary parameters, marginalizing over uncertainty in the noise. We compare the sequential estimation method and the marginalized method on real GW events from GWTC-2 using both the wavelet- and template-based models in BayesWave. We find that the recovered signals and posterior parameter distributions agree in median and width. At current sensitivities, PSD uncertainty is a subdominant effect compared to other sources of uncertainty.

Files in Document:

Other Files:

LIGO Document T2100195-v6