Relatore
Adam Sykulski - University of Leicester
Abstract
In this talk I will discuss models, inference, and applications of spatiotemporal statistics. I will focus on a spatiotemporal datatype known as trajectories (or as “Lagrangian data” to the physicist). The idea is that an object of interest is tracked as it moves through time and space to create a trajectory. In fact, many of you might have produced such datasets if you have used your phone to navigate your car or record your cycle ride. I will discuss how we can build realistic stochastic process models to understand such data, and then efficiently estimate their parameters in the spectral domain using a bias-corrected form of the Whittle Likelihood. I will demonstrate the insight gained from our methods by applying them to hundreds of millions of oceanic trajectory observations.
Organizzatore
Alessandra Luati