Modelling and robustness issues in Bayesian time series analysis
Mike West · Lecture notes-monograph series · 1996
. Some areas of recent development and current interest in time series are noted, with some discussion of Bayesian modelling efforts motivated by substantial practical problems. The areas include non-linear auto-regressive time series modelling, measurement error structures in statespace modelling of time series, and issues of timing uncertainties and time deformations. Some discussion of the needs and opportunities for work on non/semi-parametric models and robustness issues is given in each context. Keywords: Mixture models; Non-normal time series; Non-linear autoregressions; Non-parametric Bayesian models; Time series outliers; Time deformations; Timing uncertainties. 0 INTRODUCTION Three areas of recent development and current interest in Bayesian time series analysis are: non- or semi-parametric models for non-linear autoregressions, and related time series structures, based on mixture models; the modelling and accommodation of measurement errors in state space models; and timing...