Characterizing interdependencies of multiple time series. Theory and applications (Q1684637)

This book is concerned with concepts and methods for characterizing interdependencies of multiple time series in the frequency domain. The main aim of the authors is to provide an introduction to an approach to analyze vector time-series interdependence, without going into the details on the (asymptotic) theoretical aspects of statistical time-series analysis; rather, the standard results are assumed and emphasis is given on a numerical determination of estimated and large-sample theoretical implications for a practical non-large sample. With this perspective in mind, the present book consists of five chapters and an appendix of technical supplements on the concepts of Hilbert space, root modification method, and the Whittle likelihood function used in the book. It is worth mentioning that at the beginning of each chapter an abstract and a keyword list is given. In the sequel, the subject of each of the five chapters is briefly presented. Chapter 1 (Introduction) gives an introduction and a brief overview of the literature on empirical causal analysis. Topics such as the relation of causality and exogeneity in the framework of a simultaneous equation; ancillarity and conditioning in relation to exogeneity; and the relation of causality and prediction improvement in empirical analyses, are discussed. To characterize the interdependent structure of a pair of two jointly second-order stationary processes Chapter 2 (The measures of one-way effect, reciprocity, and association) initially discusses the formal relation of the Granger and Sims concept of causality and then introduces the measures of one-way effect (Sections 2.3 and 2.4), reciprocity and association (Section 2.5). The measures of interdependence are defined overall as well as frequency-wise quantities in the frequency domain, providing three ways of derivation of the frequency-wise measure. The first one is based on direct canonical factorization of the spectral density matrix, while the other two are based on distributed-lag representation and innovation orthogonalization. Chapter 3 (Representation of the partial measures) extends the measures introduced in Chapter 2 to partial measures in the presence of third-series involvement. In this frame, a partial version of the measures of interdependence in order to address this problem is introduced and studied. Explicit representations of those partial measures are given, while it is shown how they are numerically evaluated by means of a canonical factorization algorithm by the first and the third author [J. Time Ser. Anal. 31, No. 4, 229--240 (2010; Zbl 1226.65022)]. Finally, Chapter 3 also shows how the theoretical framework for stationary processes is extended to cointegrated processes. Using the stationary vector ARMA process, Chapter 4 (Inference based on the vector autoregressive and moving average model) discusses the statistical inference on the partial measures of interdependence using the standard asymptotic theory of the Whittle likelihood inference for stationary multivariate ARMA processes. The point is the use of a simulation-based estimation of the covariance matrix of the measure-related statistics. In Section 4.2, the performance of a small sample of the partial one-way effect measure estimates is evaluated using Monte Carlo data generated by a pair of trivariate data generating processes, the VAR(2) and VARMA(1,1) models. Finally, an analysis of interdependence in the frequency domain is illustrated in Section 4.3, based on US interest rates and economic growth data. Chapter 5 (Inference on changes in interdependence measures) considers the association between the structural change in parameters and changes in the causal measure. In this context, Section 5.2 provides a test for measure changes using a sub-sample variance estimation for high-frequency data. Section 5.3 investigates by Monte Carlo simulation how the proposed test approach works for small sample examples. Finally, two illustrative empirical applications are given in Section 5.4. The first application is the predictability of aggregate stock returns using dividend yields, while the second one is concerned with stock index and its futures. Summarizing, the authors have written an interesting and high valuable book, with emphasis on the practical analysis of time series. At the end of each chapter, a short list of references is provided and this will help a reader wishing to pursue this area further.