|
Abstract:
|
Environmental studies often include multivariate count processes characterized by interactions at different spatial and temporal scales, as well as, excess zeros resulting from the underlying natural phenomena and/or imperfect detection and sampling procedures. However, lack of multivariate discrete distributions that allow for complex dependence structures and excess zeros, and computational difficulties involved in application of such distributions in a joint modeling framework, necessitate the development of efficiently parameterized hierarchical Bayesian models. Furthermore, a semiparametric modeling approach provides a more realistic framework in which to account for possible non-linear relationships among variables in real-world situations. We develop hierarchical Bayesian semiparametric zero-inflated models for multivariate spatio-temporal environmental count processes.
|
