Special Issue of the Journal of Time Series Analysis in Honor of Professor Masanobu Taniguchi
本特刊收录12篇前沿论文，涵盖高维误差变量回归、函数系数协整检验、多元时间序列聚类、时空过程变点检测等主题，适合时间序列与计量经济学研究者参考。
Taniguchi Sensei – our colleague and friend Masanobu Taniguchi – retired from Waseda University in Tokyo at the end of March 2022 after a long and productive career that put Waseda on the international map of time series analysis and mathematical statistics. Masanobu arrived at Waseda from Osaka some 20 years ago and rapidly developed a powerful team of students (in total 19 theses defended) and researchers, as well as an impressive network of international collaborations. Thanks to him and the countless international conferences and symposiums he tirelessly organized all over Japan, numerous statisticians from all continents enjoyed his warm hospitality, established fruitful collaborative contacts with his team, and discovered the refinements of Japanese lifestyle and culture. Statistical inference for stochastic processes and time series is a red thread running through Masanobu's entire research career. This does not mean, however, that his contributions are narrowly concentrated on one single subject! Quite on the contrary, his scientific interests are embracing an exceptionally wide spectrum of mathematical and applied statistics topics. While it is not possible here to do justice to all of his contributions, let us mention higher-order asymptotics, a notoriously difficult subject where he can be considered to be a worldwide expert, spectral methods, local asymptotic normality and Le Cam's asymptotic theory of statistical experiments, Edgeworth expansions in stationary processes, estimating functions, discriminant analysis and clustering, empirical likelihood methods, long-memory processes, heavy tails, volatility models, … not to forget economic and financial applications, risk analysis, and portfolio theory – all in the general framework of serially dependent observations. That activity has resulted in over 150 articles published in internationally acclaimed journals including the Annals of Statistics, the Journal of the Royal Statistical Society, the Journal of the American Statistical Association, Biometrika, the Journal of Econometrics, the Journal of Time Series Analysis, Econometric Theory, the Journal of Multivariate Analysis, among many others, and no less than seven books. It is an honor for us to guest-edit this special issue of the Journal of Time Series Analysis as a tribute to Masanobu's scientific achievement. This issue contains 12 invited papers, all lying at the frontier in time series analysis research, by econometricians and statisticians. All papers were refereed as per the standards of the journal. Bhattacharjee, Chakraborty and Koul discuss the estimation of the regression parameters in a high-dimensional errors in variables linear regression model, where the measurement errors in the covariates are assumed to form a stationary short-memory moving average process having known Laplace stationary distribution and the regression errors are assumed to be independent nonidentically distributed. They also derive Massart's inequality for independent and short-memory moving average predictors. Chan and Dai deal with constant parameters testing problem in semi-parametric functional coefficient cointegrated framework. They propose an orthogonal series approximation-based test statistic to tackle the problem, and study its asymptotic theory. The proposed test is illustrated by Monte Carlo simulation and a real data analysis. Davis, Fernandes and Fokianos propose a novel methodology for clustering multivariate time series data, on the basis of energy distance. After establishing some asymptotic properties of empirical energy distance statistics for time series under a condition of α -mixing rate functions, they illustrate the proposed clustering procedure by a Monte Carlo simulation and two real data analyses. Dette and Quanz propose some asymptotically distribution-free tests of the existence of a change in the sequence of mean functions of a given spatiotemporal process, where the change occurs with a norm exceeding a given threshold. The proposed tests are based on the cumulative sum paradigm. Francq and Zakoïan develop Godambe's optimal quasi-likelihood estimating equation approach for a class of time series models that are determined by the parametrization of the first two conditional moments only. The results are illustrated by a Monte Carlo simulation and some real financial data analyses. Giraitis and Marotta deal with the problem of estimating the mean of a stationary time series observed at unevenly spaced time points. They show that any unevenly spaced sample can be used to estimate the mean of an underlying stationary linear time series. They provide an expression for the sample mean estimator and establish its asymptotic properties and the central limit theorem. Hallin, Nisol and Tavakoli are considering high-dimensional time series of functional data. Their contribution consists of two parts: the first part, under functional versions of the traditional assumptions in classical dynamic factor models, establishes a representation theorem by which a high-dimensional time series decomposes into the sum of a common component loading scalar common shocks via functional loadings and an idiosyncratic component. The second part provides consistent methods for the identification of the number of factors and the estimation of the loadings and the shocks. Hsu, Sim and Tsay discuss the problem of testing symmetry in the cross-correlation matrices of a high-dimensional stochastic process implied by exact factor models. Both simulations and real examples are used to demonstrate the applications and finite-sample performance of the proposed tests. Lee and Jo examine a first-order bivariate random coefficient integer-valued autoregressive model and investigate some of the corresponding inferential procedures: parameter estimation (conditional least squares, modified quasi-likelihood, and exponential family quasi-likelihood methods) and parameter change testing. Peña and Tsay consider the problem of clustering stationary scalar time series using their marginal properties and a hierarchical method. They propose a new test statistic for detecting whether a data set consists of multiple clusters and a new procedure to determine the number of these clusters. Their method is based on the jumps, that is, the increments, in the heights of the dendrogram when hierarchical clustering is applied to the data. Tjøstheim, Jullum and Løland give a review of some recent developments on embeddings of time series and dynamic networks. They highlight differences between the static and dynamic cases and propose several open problems in the latter case. We are grateful to all these contributors for presenting interesting new results in the research domains in which Masanobu has made influential contributions. To conclude, we also would like to express our deep appreciation to the Editor-in-Chief of the Journal of Time Series Analysis, Rob Taylor, for welcoming this project of a special issue and helping us achieve it. We give our special thanks to the anonymous referees for their careful and insightful reports, and are grateful to Priscilla Goldby for her continued support during the peer review process.