non-Gaussian Time Series Research Articles

Characterizing Nonlinear Heartbeat Dynamics Within a Point Process Framework

Human heartbeat intervals are known to have nonlinear and nonstationary dynamics. In this paper, we propose a model of R-R interval dynamics based on a nonlinear Volterra-Wiener expansion within a point process framework. Inclusion of second-order nonlinearities into the heartbeat model allows us to estimate instantaneous heart rate (HR) and heart rate variability (HRV) indexes, as well as the dynamic bispectrum characterizing higher order statistics of the nonstationary non-gaussian time series. The proposed point process probability heartbeat interval model was tested with synthetic simulations and two experimental heartbeat interval datasets. Results show that our model is useful in characterizing and tracking the inherent nonlinearity of heartbeat dynamics. As a feature, the fine temporal resolution allows us to compute instantaneous nonlinearity indexes, thus sidestepping the uneven spacing problem. In comparison to other nonlinear modeling approaches, the point process probability model is useful in revealing nonlinear heartbeat dynamics at a fine timescale and with only short duration recordings.

On the Entropy Production of Time Series with Unidirectional Linearity

There are non-Gaussian time series that admit a causal linear autoregressive moving average (ARMA) model when regressing the future on the past, but not when regressing the past on the future. The reason is that, in the latter case, the regression residuals are only uncorrelated but not statistically independent of the future. In previous work, we have experimentally verified that many empirical time series indeed show such a time inversion asymmetry. For various physical systems, it is known that time-inversion asymmetries are linked to the thermodynamic entropy production in non-equilibrium states. Here we show that such a link also exists for the above unidirectional linearity. We study the dynamical evolution of a physical toy system with linear coupling to an infinite environment and show that the linearity of the dynamics is inherited to the forward-time conditional probabilities, but not to the backward-time conditionals. The reason for this asymmetry between past and future is that the environment permanently provides particles that are in a product state before they interact with the system, but show statistical dependencies afterwards. From a coarse-grained perspective, the interaction thus generates entropy. We quantitatively relate the strength of the non-linearity of the backward conditionals to the minimal amount of entropy generation.

Improved bispectrum based tests for Gaussianity and linearity

The classical bispectrum based tests for linearity of time series are based on Gaussian asymptotics and a suboptimal smoothing in the bispectral domain. We show that the resulting classical tests may lead to vastly incorrect significance levels for non-Gaussian time series. This implies that a non-Gaussian linear time series may incorrectly be classified as non-linear. The purpose of this paper is to propose simple yet accurate tests for Gaussianity and linearity. The improved tests are derived through: (1) an optimal hexagonal smoothing in the bispectral domain, (2) the construction of simple and intuitive bispectrum based test statistics, and (3) determination of correct significance levels through a new skewness preserving scheme for linear surrogate data. The superiority of the proposed tests is demonstrated through extensive Monte Carlo simulations using relevant synthetic data.

Log-amplitude statistics of intermittent and non-Gaussian time series

In studies of hydrodynamic turbulence and nonequilibrium systems, it has been demonstrated that the observed non-Gaussian probability density functions are often described effectively by a superposition of Gaussian distributions with fluctuating variances. Based on this framework, we propose a general method to characterize intermittent and non-Gaussian time series. In our approach, an observed time series is assumed to be described by the multiplication of Gaussian and amplitude random variables, where the amplitude variable describes the variance fluctuation. It is shown analytically that statistical properties of the log-amplitude fluctuations can be estimated using the logarithmic-absolute moments of the observed time series. This method is applicable to a wide variety of symmetric unimodal distributions with heavy tails in order to quantify the deviation from a Gaussian distribution. By analyzing random cascade-type processes and superstatistical non-Gaussian models with power-law tails, we demonstrate that our method can provide detailed characterization in a wide range of non-Gaussian fluctuations.

Autoregression with Non-Gaussian Innovations

Many economics and finance time series are non-Gaussian. In this paper, we propose a Bayesian approach to non-Gaussian autoregressive time series models via quantile functions. This approach is parametric, so we also compare the proposed parametric approach with a semi-parametric approach. Simulation studies and applications to real time series show that this method works very well.

Noncausal Vector Autoregression

In this paper, we propose a new noncausal vector autoregressive (VAR) model for non-Gaussian time series. The assumption of non-Gaussianity is needed for reasons of identifiability. Assuming that the error distribution belongs to a fairly general class of elliptical distributions, we develop an asymptotic theory of maximum likelihood estimation and statistical inference. We argue that allowing for noncausality is of importance in empirical economic research, which currently uses only conventional causal VAR models. Indeed, if noncausality is incorrectly ignored, the use of a causal VAR model may yield suboptimal forecasts and misleading economic interpretations. This is emphasized in the paper by noting that noncausality is closely related to the notion of nonfundamentalness, under which structural economic shocks cannot be recovered from an estimated causal VAR model. As detecting nonfundamentalness is therefore of great importance, we propose a procedure for discriminating between causality and noncausality that can be seen as a test of nonfundamentalness. The methods are illustrated with applications to fiscal foresight and the term structure of interest rates.

Autoregressive processes with normal-Laplace marginals

The normal-Laplace distribution introduced by Reed and Jorgensen [Reed, W.J., Jorgensen, M., 2004. The double Pareto-lognormal distribution, A new parametric model for size distributions. Comm. Statist-Theory and Methods, 33, 1733–1753], is studied briefly. A first order autoregressive process with normal-Laplace stationary marginal distribution is introduced and various properties are discussed. The process gives a combination of Gaussian as well as non-Gaussian time series models for the first time and is free from the zero defect problem. Simulation studies are done and sample path properties are explored. Regression behaviour of the process is studied. The applications in modelling data from various contexts are also discussed.

Detecting ARCH Effects in Non-Gaussian Time Series

Engles ARCH test has become the standard test for ARCH effects in applied work. Under non-normality the true rejection probability of this test can differ substantially from the nominal level, however. Bootstrap and Monte Carlo versions of the test may then be used instead. This paper proposes an alternative test procedure. The new test exploits the empirical distribution of the data and an extended probability integral transformation. The test is compared with the former tests in Monte Carlo experiments. Under normality, the new test works as well as the conventional Monte Carlo test and the bootstrap. Under non-normality, the test tends to be more accurate and more powerful than the bootstrapped ARCH test. The procedure is then used to test for ARCH effects in S&P 500 returns sampled at different frequencies. In contrast to the standard and the bootstrapped ARCH tests, the new test detects ARCH effects in the transformed low-frequency returns.

Outliers Detection in Multivariate Time Series by Independent Component Analysis

In multivariate time series, outlying data may be often observed that do not fit the common pattern. Occurrences of outliers are unpredictable events that may severely distort the analysis of the multivariate time series. For instance, model building, seasonality assessment, and forecasting may be seriously affected by undetected outliers. The structure dependence of the multivariate time series gives rise to the well-known smearing and masking phenomena that prevent using most outliers' identification techniques. It may be noticed, however, that a convenient way for representing multiple outliers consists of superimposing a deterministic disturbance to a gaussian multivariate time series. Then outliers may be modeled as nongaussian time series components. Independent component analysis is a recently developed tool that is likely to be able to extract possible outlier patterns. In practice, independent component analysis may be used to analyze multivariate observable time series and separate regular and outlying unobservable components. In the factor models framework too, it is shown that independent component analysis is a useful tool for detection of outliers in multivariate time series. Some algorithms that perform independent component analysis are compared. It has been found that all algorithms are effective in detecting various types of outliers, such as patches, level shifts, and isolated outliers, even at the beginning or the end of the stretch of observations. Also, there is no appreciable difference in the ability of different algorithms to display the outlying observations pattern.

Stationary mixture transition distribution (MTD) models via predictive distributions

This paper combines two ideas to construct autoregressive processes of arbitrary order. The first idea is the construction of first order stationary processes described in Pitt et al. [(2002). Constructing first order autoregressive models via latent processes. Scand. J. Statist. 29, 657–663] and the second idea is the construction of higher order processes described in Raftery [(1985). A model for high order Markov chains. J. Roy. Statist. Soc. B. 47, 528–539]. The resulting models provide appealing alternatives to model non-linear and non-Gaussian time series.

A flexible class of parametric transition regression models based on copulas: application to poliomyelitis incidence

This paper presents an extension of a general parametric class of transitional models of order p. In these models, the conditional distribution of the current observation, given the present and past history, is a mixture of conditional distributions, each of them corresponding to the current observation, given each one of the p-lagged observations. Such conditional distributions are constructed using bivariate copula models which allow for a rich range of dependence suitable to model non-Gaussian time series. Fixed and time varying covariates can be included in the models. These models have the advantage of straightforward construction and estimation for the analysis of time series and more general longitudinal data. A poliomyelitis incidence data set is used to illustrate the proposed methods, contrary to other researches' conclusions whose methods are mainly based on linear models, we find significant evidence of a decreasing trend in polio infection after accounting for seasonality.

Detecting Autoregressive Conditional Heteroskedasticity in Non-Gaussian Time Series

In economic time series conditional heteroskedasticity and conditional non-normality may occur simultaneously. Well known examples include time series of financial returns. The present paper examines a new test for (generalized) autoregressive conditional heteroskedasticity in Monte Carlo experiments with normal, fat-tailed and/or skewed conditional distributions. In the experiments the size of the new test is accurate and the test has good power against the considered ARCH and GARCH alternatives. Under conditional normality the test is as powerful as the standard Lagrange multiplier test for ARCH effects and a robust ARCH test. Under conditional non-normality the new test has often substantially more power then the other two tests.

Periodic Markov switching autoregressive models for Bayesian analysis and forecasting of air pollution

Markov switching autoregressive models (MSARMs) are efficient tools to analyse nonlinear and non-Gaussian time series. A special MSARM with two harmonic components is proposed to analyse periodic time series. We present a full Bayesian analysis based on a Gibbs sampling algorithm for model choice and the estimations of the unknown parameters, missing data and predictive distributions. The implementation and modelling steps are developed by tackling the problem of the hidden states labeling by means of random permutation sampling and constrained permutation sampling. We apply MSARMs to study a data set about air pollution that presents periodicities since the hourly mean concentration of carbon monoxide varies according to the dynamics of the 24 day-hours and of the year. Hence, we introduce in the model both a hidden state-dependent daily component and a state-independent yearly component, giving rise to periodic MSARMs.

A multi-move sampler for estimating non-Gaussian time series models: Comments on Shephard & Pitt (1997)

This note points out a problem in the multi-move sampler as proposed by Shephard & Pitt (1997) and provides an alternative correct formulation.

비정규 시계열 자료의 회귀모형 연구

본 연구에서는 비정규 시계열 자료에 관한 다양한 회귀모형을 고찰하고, 이들 모형의 선택 기준에 관하여 연구해 보았다. 모형 선택의 기준으로는 AIC (Akaike information criterion), BIC (Baysian information criterion) 그리고 우도비 검정을 확장 적용하였다. 또한, 실제의 Polio 자료분석을 통해 이를 적용해보았다. In this paper we reviewed a variety of non-Gaussian time series models, and studied the model selection criteria such as AIC and BIC to select proper models. We also considered the likelihood ratio test and applied it to analysis of Polio data set.

Generalized Autoregressive Moving Average Models

A class of generalized autoregressive moving average (GARMA) models is developed that extends the univariate Gaussian ARMA time series model to a flexible observation-driven model for non-Gaussian time series data. The dependent variable is assumed to have a conditional exponential family distribution given the past history of the process. The model estimation is carried out using an iteratively reweighted least squares algorithm. Properties of the model, including stationarity and marginal moments, are either derived explicitly or investigated using Monte Carlo simulation. The relationship of the GARMA model to other models is shown, including the autoregressive models of Zeger and Qaqish, the moving average models of Li, and the reparameterized generalized autoregressive conditional heteroscedastic GARCH model (providing the formula for its fourth marginal moment not previously derived). The model is demonstrated by the application of the GARMA model with a negative binomial conditional distribution to a well-known time series dataset of poliomyelitis counts.

The Mixture Transition Distribution Model for High-Order Markov Chains and Non-Gaussian Time Series

The mixture transition distribution model (MTD) was introduced in 1985 by Raftery for the modeling of high-order Markov chains with a finite state space. Since then it has been generalized and successfully applied to a range of situations, including the analysis of wind directions, DNA sequences and social behavior. Here we review the MTD model and the developments since 1985. We first introduce the basic principle and then we present several extensions, including general state spaces and spatial statistics. Following that, we review methods for estimating the model parameters. Finally, a review of different types of applications shows the practical interest of the MTD model.

A distribution-free method for forecasting non-gaussian time series

The non-linear instantaneous transformation is a method to model and forecast non-Gaussian time series. A restriction of this method is that the marginal distribution of data must be known, or a general distribution form has to be determined. A difficulty of this method is that in practice the distribution of observed data is usually unknown, and it needs to be determined by fitting the data. In this study, a distribution-free plotting position formula is applied to the non-linear instantaneous transformation method. Synthetic time series and observed data are used to illustrate the proposed method, which does not require fitting the marginal distribution of the data to be forecasted.

Trispectrum deconvolution of linear processes with randomly missing observations

In a large application area of time series analysis - geophysical exploration - the underlying innovations sequence is of primary interest and must be estimated. This sequence is estimated by deconvolving the non-Gaussian stationary time series. The deconvolution of a non-Gaussian non-minimum phase linear process, when some observations are missing according to a point binomial distribution is considered. This analysis tools are the bispectrum and the trispectrum. A Monte Carlo study is performed to illustrate the proposed methods.

AN ANALYSIS OF DAILY PEAK STREAM DISCHARGE USING A NON-GAUSSIAN TIME SERIES MODEL

Daily peak stream discharge data, collected over time, are typically characterized by a few large peaks separated by runs of small values, where peaks correspond to the occurrence of storms. Furthermore, the peak discharge on the first day of a storm has little or no relationship to the previous day's discharge. These characteristics are not present in standard Gaussian time series models in which a zig-zag behavior not conducive to runs of small values is observed and the present value always depends on the previous value. However, they can be successfully captured with non-Gaussian time series models.