A0461
Title: Multi-order circular Markov processes with canonical vine representations
Authors: Hiroaki Ogata - Tokyo Metropolitan University (Japan) [presenting]
Abstract: A way of construction of multi-order circular Markov processes is introduced. Bivariate circular distributions with specified marginal distributions were previously proposed, and its representation naturally induces a circular Markov process with two arbitrary circular densities. One of the densities is called a binding density and it can be regarded as a copula density. This circular Markov process is extended to a multi-order one by employing a pair-copula decomposition of a multivariate distribution. The way of decomposition corresponds to the graphical model denoted as the canonical vine. Fitting the multi-order circular Markov process to real circular data is also considered.
