B0865
Title: General correlated statistical count structures
Authors: Robert Lund - The University of California, Santa Cruz (United States) [presenting]
Abstract: Methods are considered capable of generating a count-valued time series, a spatial random field, or a spatio-temporal random process having any prescribed marginal distribution. A Gaussian copula is used to transform a correlated Gaussian process into the desired count structure. The methods are shown to have the most general autocovariance structure achievable, permit any marginal distribution whatsoever, and can easily accommodate covariates. Hermite expansions are used to relate the autocovariance of the Gaussian process to that of the count process. Particle filtering methods of likelihood evaluation are explored.
