B1648
Title: A class of hypothesis tests for general order Markov chains based on phi-divergence
Authors: Vlad Stefan Barbu - University of Rouen-Normandy (France) [presenting]
Thomas Gkelsinis - University of Rouen - Normandy (France)
Abstract: The focus is on the new methodological contributions for assessing the fit (homogeneity and goodness-of-fit) of general order Markov chains by taking into account prior information related to the utility of each transition of the multistate system. The underlying mechanism is based on the concept of divergence measure and, particularly, on the family of weighted phi-divergences between general order Markov chains, with special cases being the chi-squared and likelihood ratio tests. Accordingly, that methodology can be seen as a broad generalization, where the existing related techniques are particular members when no prior information is considered. The corresponding asymptotic theory is presented with Monte Carlo simulations for evaluating the performance of the proposed methodology.
