Abstract The local linear regression technique is applied to estimation of functional-coefficient models for time series data. include threshold autoregressive and as special cases but with the added advantages such depicting finer structure underlying dynamics better postsample forecasting performance. Also proposed are a new bootstrap test goodness fit bandwidth selector based on newly defined cross-validatory expected errors. methodology data-analytic sufficient flexibility analyze...

Journal Article Efficient estimation of conditional variance functions in stochastic regression Get access JIANQING FAN, FAN Department Statistics, University CaliforniaLos Angeles, California 90095, U.S.A.jfan@stat.cla.edu Search for other works by this author on: Oxford Academic Google Scholar QIWEI YAO Institute Mathematics and KentCanterbury, Kent CT2 7NF, U.K.q.yao@ukc.ac.uk Biometrika, Volume 85, Issue 3, September 1998, Pages 645–660, https://doi.org/10.1093/biomet/85.3.645 Published:...

This paper deals with the factor modeling for high-dimensional time series based on a dimension-reduction viewpoint. Under stationary settings, inference is simple in sense that both number of factors and loadings are estimated terms an eigenanalysis nonnegative definite matrix, therefore applicable when dimension order few thousands. Asymptotic properties proposed method investigated under two settings: (i) sample size goes to infinity while fixed; (ii) go together. In particular, our...

ARCH and GARCH models directly address the dependency of conditional second moments, have proved particularly valuable in modelling processes where a relatively large degree fluctuation is present. These include financial time series, which can be heavy tailed. However, little known about properties or heavy–tailed setting, no methods are available for approximating distributions parameter estimators there. In this paper we show that, errors, asymptotic quasi–maximum likelihood nonnormal,...

Abstract Motivated by the problem of setting prediction intervals in time series analysis, we suggest two new methods for conditional distribution estimation. The first method is based on locally fitting a logistic model and spirit recent work parametric techniques density It produces estimators that may be arbitrarily high order but nevertheless always lie between 0 1. second involves an adjusted form Nadaraya–Watson estimator. preserves bias variance properties class second-order...

Summary Varying-coefficient linear models arise from multivariate nonparametric regression, non-linear time series modelling and forecasting, functional data analysis, longitudinal analysis others. It has been a common practice to assume that the varying coefficients are functions of given variable, which is often called an index. To enlarge capacity substantially, this paper explores class varying-coefficient in index unknown estimated as combination regressors and/or other variables. We...

Abstract Motivated by the problem of setting prediction intervals in time series analysis, we suggest two new methods for conditional distribution estimation. The first method is based on locally fitting a logistic model and spirit recent work parametric techniques density It produces estimators that may be arbitrarily high order but nevertheless always lie between 0 1. second involves an adjusted form Nadaraya–Watson estimator. preserves bias variance properties class second-order...

Journal Article Estimation of latent factors for high-dimensional time series Get access Clifford Lam, Lam Department Statistics, London School Economics, WC2A 2AE, U.K.c.lam2@lse.ac.uk Search other works by this author on: Oxford Academic Google Scholar Qiwei Yao, Yao U.K.q.yao@lse.ac.uk Neil Bathia Jump Trading, 600 West Chicago Avenue, Chicago, Illinois 60654, U.S.A.nbathia@jumptrading.com Biometrika, Volume 98, Issue 4, December 2011, Pages 901–918, https://doi.org/10.1093/biomet/asr048...

This paper considers the nonparametric estimation of regression expectiles and percentiles by using an asymmetric least squares (ALS) approach, in which squared error loss function is given different weights depending on whether thc residual positive or negative. The kernel method based locally linear fit adopted, also provides estimator derivative function. Under assumption that observations are strictly stationary ρ-mixing asymptotic normality for estimators conditional established...

We suggest two improved methods for conditional density estimation. The first is based on locally fitting a log-linear model, and in the spirit of recent work parametric techniques second method constrained local polynomial estimator. Both always produce non-negative estimators. propose an algorithm suitable selecting bandwidths either also develop new bootstrap test symmetry functions. proposed are illustrated by both simulation application to real data set.

We propose a new method for estimating common factors of multiple time series. One distinctive feature the approach is that it applicable to some nonstationary The unobservable, are identified by expanding white noise space step step, thereby solving high-dimensional optimization problem several low-dimensional sub-problems. Asymptotic properties estimation investigated. proposed methodology illustrated with both simulated and real datasets.

We propose a hybrid approach for the modeling and short-term forecasting of electricity loads. Two building blocks our are (1) overall trend seasonality by fitting generalized additive model to weekly averages load (2) dependence structure across consecutive daily loads via curve linear regression. For latter, new methodology is proposed regression with both response regressors. The key idea behind dimension reduction based on singular value decomposition in Hilbert space, which reduces...

It is increasingly important in financial economics to estimate volatilities of asset returns. However, most the available methods are not directly applicable when number assets involved large, due lack accuracy estimating high-dimensional matrices. Therefore it pertinent reduce effective size volatility matrices order produce adequate estimates and forecasts. Furthermore, since high-frequency data for different typically recorded at same time points, conventional dimension-reduction...

We consider a class of vector autoregressive models with banded coefficient matrices. This setting represents type sparse structure for high-dimensional time series, although the implied auto-covariance matrices are not banded. The is also practically meaningful when component series ordered appropriately. establish convergence rates estimated propose Bayesian information criterion determining width bands in matrices, which proved to be consistent. By exploring some approximate structures...

The local linear regression technique is applied to estimation of functional-coefficient models for time series data. include threshold autoregressive and as special cases but with the added advantages such depicting finer structure underlying dynamics better postsample forecasting performance. Also proposed are a new bootstrap test goodness fit bandwidth selector based on newly defined cross-validatory expected errors. methodology data-analytic sufficient flexibility analyze complex...

AbstractIn the analysis of microarray data, and in some other contemporary statistical problems, it is not uncommon to apply hypothesis tests a highly simultaneous way. The number, N say, used can be much larger than sample sizes, n, which are applied, yet we wish calibrate so that overall level test accurate. Often sampling distribution quite different for each test, there may an opportunity combine data across samples. In this setting, how large be, as function before accuracy becomes...

We propose a new and easy-to-use method for identifying cointegrated components of nonstationary time series, consisting an eigenanalysis certain nonnegative definite matrix. Our setting is model-free, we allow the integer-valued integration orders observable series to be unknown, possibly differ. Consistency estimates cointegration space rank established both when dimension fixed as sample size increases, it diverges slowly. The proposed methodology also extended justified in fractional...

We propose a new omnibus test for vector white noise using the maximum absolute autocorrelations and cross-correlations of component series. Based on an approximation by |$L_\infty$|-norm normal random vector, critical value can be evaluated bootstrapping from multivariate distribution. In contrast to conventional test, method is proved valid testing departure that not independent identically distributed. illustrate accuracy power proposed simulation, which also shows outperforms several...

Abstract We consider to model matrix time series based on a tensor canonical polyadic (CP)-decomposition. Instead of using an iterative algorithm which is the standard practice for estimating CP-decompositions, we propose new and one-pass estimation procedure generalized eigenanalysis constructed from serial dependence structure underlying process. To overcome intricacy solving rank-reduced eigenequation, further refined approach projects it into lower-dimensional full-ranked eigenequation....

The purpose of this paper is to discuss the tests detect an epidemic alternative in mean value a sequence independent normal variables. Various test statistics, such as likelihood ratio, score-like statistic, Levin & Kline's semilikelihood and recursive residual, are studied. large deviation approximations significance levels powers developed by integrating conditional boundary crossing probabilities. Some results Monte Carlo experiments confirm accuracy these approximations. A numerical...