Time Series Analysis

by

James Douglas Hamilton

This is a large text in time series analysis that is designed for graduate students as the author acknowledges in his preface. It deals primarily with the theory and the tools rather than providing practical applications. It does not require a Ph.D. but does require a fair amount of mathematical sophistication that comes from advanced courses in probability and statistics. There are many good books at this level. This one has some unique features. It covers the traditional ARIMA models that can be found in most texts and uses the operator notation that Box and Jenkins introduced. It adds vector autoregressions which is fairly recent material. Spectral analysis (the frequency domain approach)is also covered and asymptotic theory is presented. Linear systems (more common to econometric time series than in the standard statistical books) is covered. Topics not commonly covered in competitor texts include nonstationary cases (both univariate and multivariate)with unit roots to the characteristic equation, Bayesian approaches, heteroscedastic models including the ARCH models and the topic of cointegration originally developed by Clive Granger. The book is loaded with references to the literature and is slanted towards methods useful in econometrics. Other good books at this level include Brockwell and Davis (1987), Fuller (1976), Anderson (1971), Harvey (1981) and Shumway and Stoffer (2000). Good texts solely in the frequency domain include Bloomfeld (1976), Priestley (1981), Koopmans (1974) and Brillinger (1981). Box, Jenkins and Reinsel (1994) provides practical applications using the Box-Jenkins time domain approach. (Statman13)

Download