Toroidal diffusion processes with exact likelihood inference
Conference
64th ISI World Statistics Congress
Format: IPS Abstract
Session: IPS 150 - Statistical inference for stochastic ordinary and partial differential equations
Monday 17 July 2 p.m. - 3:40 p.m. (Canada/Eastern)
Abstract
A new class of diffusion processes on the multivariate torus is presented. The aim is to model time series of angular data. The processes have explicit transition probability densities, which enables exact likelihood inference. The diffusions are ergodic and time-reversible and can be constructed for any pre-specified stationary distribution on the torus. Asymptotic likelihood theory is presented, and it is shown how exact diffusion bridge simulation can easily be done for these models. A class of circular jump processes with similar properties is proposed too. Examples illustrate the methodology.