The development of sophisticated computational methods for inference and forecasting has resulted in an enormous increase of the time series literature in recent years. This book integrates existing and new developments in time series analysis by presenting an overview of the main models and methods for inference and forecasting from the time-domain and frequency-domain perspectives, exploring connections between these two approaches. The book focuses on maximum likelihood and Bayesian methodologies. An overview of the key simulation-based methods for model fitting and prediction, including Gibbs sampling, general MCMC methods and particle filters is provided. The book is divided in two parts. The first part deals with the basic univariate time series models, inference and forecasting methods. Traditional time series models such as AR and ARMA models, dynamic linear models and general state-space models are discussed. Other non-linear and non-Gaussian univariate models such as ARCH, stochastic volatility and related models are also included in Part I. Part II is dedicated to more advanced models and core methodology in the analysis of multiple and multivariate time series.The time series models and theory are illustrated by considering data analyses from a wide range of application areas, such as signal procesing, finance and the earth and life sciences.
This book is an important reference on state-of-the-art time series methodology, serving as a guide to recent literature and encouraging readers to further investigate areas of their particular interest by pointing them to the appropriate references.
