Speaker
Description
LISA's data will have several non-stationary stochastic components: several contributions to the instrumental noise directly depend on the arm lengths of the constellation, while others will be only quasi-stationary with slow drifts over time.
Additionally, anisotropic SGWBs will show an induced non-stationarity as LISA orbits the sun. This effect will be particularly strong for the galactic foreground, but may also be present in boosted isotropic cosmological SGWBs.
All of these effects combine to create a challenging non-stationary environment for an SGWB search. In the Fourier domain, this non-stationarity will induce correlations between neighboring frequencies, as well as overestimate the variance of the data in time periods with lower-than-average noise. The Wilson-Debauchies-Meyer (WDM) wavelet domain is a complete basis time-frequency representation of the data, and allows for uncorrelated basis elements in quasi-stationary data as well as tracking of the variance over time. It also has several other technical advantages over other time-frequency representations.
In this talk, I will present a WDM-based SGWB pipeline developed as a global fit module for both GLASS and Erebor. It shows improved inference for cosmological signals in the presence of non-stationary noise components, and improves inference on discrete signals as well, since it is a better whitening filter for arm-varying LISA data.