Contributed Talk - Splinter EScience
Probabilistic Cross-Correlation for Delay Estimation
Nikos Gianniotis, Francisco Pozo Nunez, Kai Polsterer
Astroinformatics, Heidelberg Institute for Theoretical Studies (HITS)
The Interpolated Cross-Correlation Function (ICCF) has been the workhorse of astronomers when estimating the delay between pairs of lightcurves originating from AGN. We present a probabilistic reformulation of ICCF that enjoys several benefits such as accounting for measurement error, out-of-sample predictions, and most importantly, the capability of delivering a posterior distribution of the delay. Our reformulation views the observed lightcurves in each band as the manifestation of a common latent signal that we model as a sample from a Gaussian process. We demonstrate the advantages of the probabilistic cross-correlation arising from its probabilistic grounding on a number of AGN objects.