-
Frequency Domain Statistical Inference for High-Dimensional Time Series
Authors:
Jonas Krampe,
Efstathios Paparoditis
Abstract:
Analyzing time series in the frequency domain enables the development of powerful tools for investigating the second-order characteristics of multivariate processes. Parameters like the spectral density matrix and its inverse, the coherence or the partial coherence, encode comprehensively the complex linear relations between the component processes of the multivariate system. In this paper, we dev…
▽ More
Analyzing time series in the frequency domain enables the development of powerful tools for investigating the second-order characteristics of multivariate processes. Parameters like the spectral density matrix and its inverse, the coherence or the partial coherence, encode comprehensively the complex linear relations between the component processes of the multivariate system. In this paper, we develop inference procedures for such parameters in a high-dimensional, time series setup. Towards this goal, we first focus on the derivation of consistent estimators of the coherence and, more importantly, of the partial coherence which possess manageable limiting distributions that are suitable for testing purposes. Statistical tests of the hypothesis that the maximum over frequencies of the coherence, respectively, of the partial coherence, do not exceed a prespecified threshold value are developed. Our approach allows for testing hypotheses for individual coherences and/or partial coherences as well as for multiple testing of large sets of such parameters. In the latter case, a consistent procedure to control the false discovery rate is developed. The finite sample performance of the inference procedures introduced is investigated by means of simulations and applications to the construction of graphical interaction models for brain connectivity based on EEG data are presented.
△ Less
Submitted 17 January, 2024; v1 submitted 5 June, 2022;
originally announced June 2022.
-
Bootstrapping Whittle Estimators
Authors:
Jens-Peter Kreiss,
Efstathios Paparoditis
Abstract:
Fitting parametric models by optimizing frequency domain objective functions is an attractive approach of parameter estimation in time series analysis. Whittle estimators are a prominent example in this context. Under weak conditions and the (realistic) assumption that the true spectral density of the underlying process does not necessarily belong to the parametric class of spectral densities fitt…
▽ More
Fitting parametric models by optimizing frequency domain objective functions is an attractive approach of parameter estimation in time series analysis. Whittle estimators are a prominent example in this context. Under weak conditions and the (realistic) assumption that the true spectral density of the underlying process does not necessarily belong to the parametric class of spectral densities fitted, the distribution of Whittle estimators typically depends on difficult to estimate characteristics of the underlying process. This makes the implementation of asymptotic results for the construction of confidence intervals or for assessing the variability of estimators, difficult in practice. This paper proposes a frequency domain bootstrap method to estimate the distribution of Whittle estimators which is asymptotically valid under assumptions that not only allow for (possible) model misspecification but also for weak dependence conditions which are satisfied by a wide range of stationary stochastic processes. Adaptions of the bootstrap procedure developed to incorporate different modifications of Whittle estimators proposed in the literature, like for instance, tapered, de-biased or boundary extended Whittle estimators, are also considered. Simulations demonstrate the capabilities of the bootstrap method proposed and its good finite sample performance. A real-life data analysis also is presented.
△ Less
Submitted 23 July, 2021;
originally announced July 2021.
-
A Frequency Domain Bootstrap for General Multivariate Stationary Processes
Authors:
Marco Meyer,
Efstathios Paparoditis
Abstract:
For many relevant statistics of multivariate time series, no valid frequency domain bootstrap procedures exist. This is mainly due to the fact that the distribution of such statistics depends on the fourth-order moment structure of the underlying process in nearly every scenario, except for some special cases like Gaussian time series. In contrast to the univariate case, even additional structural…
▽ More
For many relevant statistics of multivariate time series, no valid frequency domain bootstrap procedures exist. This is mainly due to the fact that the distribution of such statistics depends on the fourth-order moment structure of the underlying process in nearly every scenario, except for some special cases like Gaussian time series. In contrast to the univariate case, even additional structural assumptions such as linearity of the multivariate process or a standardization of the statistic of interest do not solve the problem. This paper focuses on integrated periodogram statistics as well as functions thereof and presents a new frequency domain bootstrap procedure for multivariate time series, the multivariate frequency domain hybrid bootstrap (MFHB), to fill this gap. Asymptotic validity of the MFHB procedure is established for general classes of periodogram-based statistics and for stationary multivariate processes satisfying rather weak dependence conditions. A simulation study is carried out which compares the finite sample performance of the MFHB with that of the moving block bootstrap.
△ Less
Submitted 3 February, 2021;
originally announced February 2021.
-
Estimation of the Distribution of the Individual Reproduction Number: The Case of the COVID-19 Pandemic
Authors:
Alexander Braumann,
Jonas Krampe,
Jens-Peter Kreiss,
Efstathios Paparoditis
Abstract:
We investigate the problem of estimating the distribution of the individual reproduction number governing the COVID-19 pandemic. Under the assumption that this random variable follows a Negative Binomial distribution, we focus on constructing estimators of the parameters of this distribution using reported infection data and taking into account issues like under-reporting or the time behavior of t…
▽ More
We investigate the problem of estimating the distribution of the individual reproduction number governing the COVID-19 pandemic. Under the assumption that this random variable follows a Negative Binomial distribution, we focus on constructing estimators of the parameters of this distribution using reported infection data and taking into account issues like under-reporting or the time behavior of the infection and of the reporting processes. To this end, we extract information from regionally dissaggregated data reported by German health authorities, in order to estimate not only the mean but also the variance of the distribution of the individual reproduction number. In contrast to the mean, the latter parameter also depends on the unknown under-reporting rate of the pandemic. The estimates obtained allow not only for a better understanding of the time-varying behavior of the expected value of the individual reproduction number but also of its dispersion, for the construction of bootstrap confidence intervals and for a discussion of the implications of different policy interventions. Our methodological investigations are accompanied by an empirical study of the development of the COVID-19 pandemic in Germany, which shows a strong overdispersion of the individual reproduction number.
△ Less
Submitted 19 January, 2021;
originally announced January 2021.
-
Structural Inference in Sparse High-Dimensional Vector Autoregressions
Authors:
Jonas Krampe,
Efstathios Paparoditis,
Carsten Trenkler
Abstract:
We consider statistical inference for impulse responses in sparse, structural high-dimensional vector autoregressive (SVAR) systems. We introduce consistent estimators of impulse responses in the high-dimensional setting and suggest valid inference procedures for the same parameters. Statistical inference in our setting is much more involved since standard procedures, like the delta-method, do not…
▽ More
We consider statistical inference for impulse responses in sparse, structural high-dimensional vector autoregressive (SVAR) systems. We introduce consistent estimators of impulse responses in the high-dimensional setting and suggest valid inference procedures for the same parameters. Statistical inference in our setting is much more involved since standard procedures, like the delta-method, do not apply. By using local projection equations, we first construct a de-sparsified version of regularized estimators of the moving average parameters associated with the VAR system. We then obtain estimators of the structural impulse responses by combining the aforementioned de-sparsified estimators with a non-regularized estimator of the contemporaneous impact matrix, also taking into account the high-dimensionality of the system. We show that the distribution of the derived estimators of structural impulse responses has a Gaussian limit. We also present a valid bootstrap procedure to estimate this distribution. Applications of the inference procedure in the construction of confidence intervals for impulse responses as well as in tests for forecast error variance decomposition are presented. Our procedure is illustrated by means of simulations.
△ Less
Submitted 2 June, 2021; v1 submitted 30 July, 2020;
originally announced July 2020.
-
Statistical Estimation of High-Dimensional Vector Autoregressive Models
Authors:
Jonas Krampe,
Efstathios Paparoditis
Abstract:
High-dimensional vector autoregressive (VAR) models are important tools for the analysis of multivariate time series. This paper focuses on high-dimensional time series and on the different regularized estimation procedures proposed for fitting sparse VAR models to such time series. Attention is paid to the different sparsity assumptions imposed on the VAR parameters and how these sparsity assumpt…
▽ More
High-dimensional vector autoregressive (VAR) models are important tools for the analysis of multivariate time series. This paper focuses on high-dimensional time series and on the different regularized estimation procedures proposed for fitting sparse VAR models to such time series. Attention is paid to the different sparsity assumptions imposed on the VAR parameters and how these sparsity assumptions are related to the particular consistency properties of the estimators established. A sparsity scheme for high-dimensional VAR models is proposed which is found to be more appropriate for the time series setting considered. Furthermore, it is shown that, under this sparsity setting, threholding extents the consistency properties of regularized estimators to a wide range of matrix norms. Among other things, this enables application of the VAR parameters estimators to different inference problems, like forecasting or estimating the second-order characteristics of the underlying VAR process. Extensive simulations compare the finite sample behavior of the different regularized estimators proposed using a variety of performance criteria.
△ Less
Submitted 9 June, 2020;
originally announced June 2020.
-
Bootstrap Based Inference for Sparse High-Dimensional Time Series Models
Authors:
J. Krampe,
J-P. Kreiss,
E. Paparoditis
Abstract:
Fitting sparse models to high-dimensional time series is an important area of statistical inference. In this paper we consider sparse vector autoregressive models and develop appropriate bootstrap methods to infer properties of such processes. Our bootstrap methodology generates pseudo time series using a model-based bootstrap procedure which involves an estimated, sparsified version of the underl…
▽ More
Fitting sparse models to high-dimensional time series is an important area of statistical inference. In this paper we consider sparse vector autoregressive models and develop appropriate bootstrap methods to infer properties of such processes. Our bootstrap methodology generates pseudo time series using a model-based bootstrap procedure which involves an estimated, sparsified version of the underlying vector autoregressive model. Inference is performed using so-called de-sparsified or de-biased estimators of the autoregressive model parameters. We derive the asymptotic distribution of such estimators in the time series context and establish asymptotic validity of the bootstrap procedure proposed for estimation and, appropriately modified, for testing purposes. In particular we focus on testing that large groups of autoregressive coefficients equal zero. Our theoretical results are complemented by simulations which investigate the finite sample performance of the bootstrap methodology proposed. A real-life data application is also presented.
△ Less
Submitted 24 September, 2019; v1 submitted 28 June, 2018;
originally announced June 2018.
-
A Frequency Domain Bootstrap for General Stationary Processes
Authors:
Marco Meyer,
Efstathios Paparoditis,
Jens-Peter Kreiss
Abstract:
Existing frequency domain methods for bootstrapping time series have a limited range. Consider for instance the class of spectral mean statistics (also called integrated periodograms) which includes many important statistics in time series analysis, such as sample autocovariances and autocorrelations among other things. Essentially, such frequency domain bootstrap procedures cover the case of line…
▽ More
Existing frequency domain methods for bootstrapping time series have a limited range. Consider for instance the class of spectral mean statistics (also called integrated periodograms) which includes many important statistics in time series analysis, such as sample autocovariances and autocorrelations among other things. Essentially, such frequency domain bootstrap procedures cover the case of linear time series with independent innovations, and some even require the time series to be Gaussian. In this paper we propose a new, frequency domain bootstrap method which is consistent for a much wider range of stationary processes and can be applied to a large class of periodogram-based statistics. It introduces a new concept of convolved periodograms of smaller samples which uses pseudo periodograms of subsamples generated in a way that correctly imitates the weak dependence structure of the periodogram. %The new bootstrap procedure %corrects for those aspects of the distribution of spectral means that cannot be mimicked by existing procedures. We show consistency for this procedure for a general class of stationary time series, ranging clearly beyond linear processes, and for general spectral means and ratio statistics. Furthermore, and for the class of spectral means, we also show, how, using this new approach, existing bootstrap methods, which replicate appropriately only the dominant part of the distribution of interest, can be corrected. The finite sample performance of the new bootstrap procedure is illustrated via simulations.
△ Less
Submitted 18 June, 2018;
originally announced June 2018.
-
Testing equality of spectral density operators for functional linear processes
Authors:
Anne Leucht,
Efstathios Paparoditis,
Theofanis Sapatinas
Abstract:
The problem of testing equality of the entire second order structure of two independent functional linear processes is considered. A fully functional $L^2$-type test is developed which evaluates, over all frequencies, the Hilbert-Schmidt distance between the estimated spectral density operators of the two processes. The asymptotic behavior of the test statistic is investigated and its limiting dis…
▽ More
The problem of testing equality of the entire second order structure of two independent functional linear processes is considered. A fully functional $L^2$-type test is developed which evaluates, over all frequencies, the Hilbert-Schmidt distance between the estimated spectral density operators of the two processes. The asymptotic behavior of the test statistic is investigated and its limiting distribution under the null hypothesis is derived. Furthermore, a novel frequency domain bootstrap method is developed which approximates more accurately the distribution of the test statistic under the null than the large sample Gaussian approximation obtained. Asymptotic validity of the bootstrap procedure is established and consistency of the bootstrap-based test under the alternative is proved. Numerical simulations show that, even for small samples, the bootstrap-based test has very good size and power behavior. An application to meteorological functional time series is also presented.
△ Less
Submitted 14 April, 2020; v1 submitted 10 April, 2018;
originally announced April 2018.
