Regression Models for Time Series Analysis

 A relatively recent statistical development is the important class of models known as  generalized linear models (GLM)

 that was introduced by Nelder and Wedderburn (1972), and which provides under some conditions

 a unified regression theory suitable for continuous, categorical, and count data. The theory of GLM was originally intended for

 independent data, but  it can be extended to dependent data under various assumptions. In  the first part of this  book  the GLM   methodology is  extended systematically to time series where the primary and covariate data are both random and stochastically dependent. There are three notions which enable this ( (Fokianos and Kedem (1998), (Slud and Kedem (1994)). The notion of an

 increasing sequence of histories relative to an observer, the notion partial likelihood introduced by Cox (1975)  and the

 notion of martingale with respect to a sequence of histories. In addition this book introduces the reader

 to various regression models developed during the last thirty years or so, particularly regression models

 for integer valued time series and  summarizes classical and more recent results concerning state space models. The last part of this book  presents a Bayesian approach to prediction and interpolation in spatial data adapted to time series that may be short   and/or observed irregularly.

Reviews

More reviews on the Publisher�s  website

Table of Contents (ps  and pdf). 

  You can also obtain a zip file with all the datasets.

Here is a list of known errata. We try to keep it updated but any comments/suggestions are welcome.

1. Pdf format
2. Ps format

30/08/2004