|
Abstract:
|
Change point problems have been studied in many different forms by econometricians, ecologists, and biologists alike. Classical change point models consider serial observations partitioned into blocks, wherein observations are assumed to come from a single probability model. This talk introduces Bayesian methodology addressing a generalized change point framework that encompasses a wide range of such scenarios. Most significantly, this framework allows the data to reside on any connected graph structure. This generalization allows for applications of change point models to spatial data. In this talk, I will describe the hierarchical Bayes model and comment on the implementation. To illustrate, I will apply the methodology to modeling New Haven housing values.
|
