Innovations algorithm asymptotics for periodically stationary time series with heavy tails

The innovations algorithm can be used to obtain parameter estimates for periodically stationary time series models. In this paper we compute the asymptotic distribution for these estimates in the case where the underlying noise sequence has infinite fourth moment but finite second moment. In this case, the sample covariances on which the innovations algorithm are based are known to be asymptotically stable. The asymptotic results developed here are useful to determine which model parameters are significant. In the process, we also compute the asymptotic distributions of least squares estimates of parameters in an autoregressive model.     

Asymptotic Estimates of Elementary Probability Distributions

Several new asymptotic estimates (with precise error bounds) are derived for Poisson and binomial distributions as the parameters tend to infinity. The analytic methods used are also applicable to other discrete distribution functions.     

Large sample properties of the mle and mcle for the natural parameter of a truncated exponential family

Summary Consider a truncated exponential family of absolutely continuous distributions with natural parameter θ and truncation parameter γ. Strong consistency and asymptotic normality are shown to hold for the maximum likelihood and maximum conditional likelihood estimates of θ with γ unknown. Moreover, these two estimates are also shown to have the same limiting distribution, coinciding with that of the maximum likelihood estimate for θ when γ is assumed to be known.     

Sequential fixed accuracy estimation for nonstationary autoregressive processes

For an autoregressive process of order p, the paper proposes new sequential estimates for the unknown parameters based on the least squares (LS) method. The sequential estimates use p stopping rules for collecting the data and presumes a special modification the sample Fisher information matrix in the LS estimates. In case of Gaussian disturbances, the proposed estimates have non-asymptotic normal joint distribution for any values of unknown autoregressive parameters. It is shown that in the i.i.d. case with unspecified error distributions, the new estimates have the property of uniform asymptotic normality for unstable autoregressive processes under some general condition on the parameters. Examples of unstable autoregressive models satisfying this condition are considered.

     

A characterization of limiting distributions of regular estimates

Summary We consider a sequence of estimates in a sequence of general estimation problems with a k-dimensional parameter. Under certain very general conditions we prove that the limiting distribution of the estimates, if properly normed, is a convolution of a certain normal distribution, which depends only of the underlying distributions, and of a further distribution, which depends on the choice of the estimate. As corollaries we obtain inequalities for asymptotic variances and for asymptotic probabilities of certain sets, generalizing so some results of J. Wolfowitz (1965), S. Kaufman (1966), L. Schmetterer (1966) and G. G. Roussas (1968).     

Random walks with non-convolution equivalent increments and their applications

This paper mainly presents some global and local asymptotic estimates for the tail probabilities of the supremum and overshoot of a random walk in “the intermediate case”, where the related distributions of the increments of the random walk may not belong to the convolution equivalent distribution class. Some of the obtained results can include the classical results. For this, the paper first introduces some new distribution classes using the γ-transform of distributions, and investigates their properties and relations with some other existing distribution classes. Based on the above results, some equivalent conditions for the global and local asymptotics of the γ-transform of the distribution of the supremum of the above random walk are given. Applying these results to risk theory and infinitely divisible laws, the paper obtains some asymptotic estimates for the ruin probability and the local ruin probability of the renewal risk model with non-convolution equivalent claims, and the global and local asymptotics of an infinitely divisible law with a non-convolution equivalent Lévy measure.     

Optimal training sample allocation and asymptotic expansions for error rates in discriminant analysis

The sample-based rule obtained from Bayes classification rule by replacing the unknown parameters by ML estimates from a stratified training sample is used for the classification of a random observationX into one ofL populations. The asymptotic expansions in terms of the inverses of the training sample sizes for cross-validation, apparent and plug-in error rates are found. These are used to compare estimation methods of the error rate for a wide range of regular distributions as probability models for considered populations. The optimal training sample allocation minimizing the asymptotic expected error regret is found in the cases of widely applicable, positively skewed distributions (Rayleigh and Maxwell distributions). These probability models for populations are often met in ecology and biology. The results indicate that equal training sample sizes for each populations sometimes are not optimal, even when prior probabilities of populations are equal.     

On the mean square error of maximum likelihood estimates of the distribution density of sufficient statistics of the multivariate normal distribution

The extensive use of maximum likelihood estimates underscores the importance of the problem of statistical estimation of their errors. These estimates are of utmost importance in cases where the family of normal distributions and the families related to the normal distributions are considered [1, 2, 4]. The mean square errors of the maximum likelihood estimates of the normal density were investigated in the author's paper [3]. The mean square errors of statistical estimates of some families of densities related to the normal distributions were considered in the papers [4–6]. In the present paper, we obtain an asymptotic expansion of the mean square error of the maximum likelihood estimates of the densities of the joint distribution of sufficient statistics of the family of multivariate normal distributions. The results obtained allow us to construct the mean square errors of the maximum likelihood estimates for the chi-square density and Wishart's density. Translated fromStatisticheskie Metody Otsenivaniya i Proverki Gipotez, pp. 4–11, Perm. 1990.     

On approximations to generalized Poisson distributions

In this paper three methods of the construction of approximations to generalized Poisson distributions are considered: approximation by a normal law, approximation by asymptotic distributions, the so-called Robbins mixtures, and approximation with the help of asymptotic expansions. Uniform and (for the first two methods) nonuniform estimates of the accuracy of the corresponding approximations are given. Some estimates for the concentration functions of generalized Poisson distributions are also presented. Supported by the Russian Foundation for Fundamental Research (grant No. 93-01-01446). Proceedings of the XVII Seminar on Stability Problems for Stochastic Models, Kazan, Russia, 1995, Part II.     

Optimum step-stress partially accelerated life tests for the truncated logistic distribution with censoring

This paper deals with the optimal designing of step-stress partially accelerated life tests (PALTs) in which items are run at both accelerated and use conditions under censoring. It is assumed that the lifetime of the items follow truncated logistic distribution truncated at point zero. Truncated distributions arise when sample selection is not possible in some sub-region of the sample space. The logistic distribution is considered inappropriate for modeling lifetime data because left hand side of its distribution extends to negative infinity, and this could conceivably result in modeling negative times-to-failure. This has necessitated the use of truncated logistic distribution truncated at point zero for modeling lifetime data. Unlike the widely studied exponential, Weibull and lognormal life distributions, the failure rate of truncated logistic distribution is increasing and more realistically bounded below and above by non-zero finite quantity. The optimal change-time for the step PALT is determined by minimizing either the generalized asymptotic variance of maximum likelihood estimates (MLEs) of the acceleration factor and the hazard rate at use condition or the asymptotic variance of MLE of the acceleration factor. Inferential procedure involving model parameters and acceleration factor are studied. Sensitivity analysis is also performed.     

Estimate for Euclidean parameters of a mixture of two symmetric distributions

A sample from a mixture of two symmetric distributions is observed. The considered distributions differ only by a shift. Estimates are constructed by the method of estimating equations for parameters of mean locations and concentrations (mixing probabilities) of both components. We obtain conditions for the asymptotic normality of these estimates. The greatest lower bounds for the coefficients of dispersion of the estimates are determined.     

On the real accuracy of approximation in the central limit theorem. II

In the present article, we obtain new explicit estimates for accuracy of approximation in the central limit theorem (CLT). We construct these approximations with the use of asymptotic expansions. We compare the estimates with the real accuracy of approximation for a specific distribution. We also discuss the following question: Why the estimate from the Berry–Esseen theorem cannot catch even the order of proximity of distributions in the CLT?     

Determining <Emphasis Type="Italic">p</Emphasis>-values for systems cointegration tests with a prior adjustment for deterministic terms

Following Doornik (J Econ Surv 12:573–593, 1998) I present a procedure to approximate the asymptotic distributions of systems cointegration tests with a prior adjustment for deterministic terms suggested by Lütkepohl (Econometrica 72:647–662, 2004), Saikkonen and Lütkepohl (Econometric Theory 16:373–406, 2000a, J Business Econ Stat 18:451–464, 2000b, Time Series Anal 21:435–456, 2000c) and Saikkonen and Luukkonen (J Econ 81:93–126, 1997). These tests rely upon different assumptions as to the inclusion of deterministic components such as a constant, a linear trend or a level shift. The asymptotic distributions, which are functions of Brownian motions, are approximated by Gamma distributions. Only estimates of the mean and variance of the asymptotic test distributions are needed to fit the Gamma distributions. Such estimates are obtained from response surfaces. The required coefficients to compute the asymptotic moments are presented in this paper. Via the fitted Gamma distributions one can, then, easily derive p-values or arbitrary percentiles.     

Data Sphering: Some Properties and Applications

Centring-then-sphering is a very important pretreatment in data analysis. The purpose of this paper is to study the asymptotic behavior of the sphering matrix based on the square root decomposition (SRD for short) and its applications. A sufficient condition is given under which SRD has nondegenerate asymptotic distribution. As examples, some commonly used and affine equivariant estimates of the dispersion matrix are shown to satisfy this condition. The case when the population dispersion matrix varies is also treated. Applications to projection pursuit (PP) are presented. It is shown that for elliptically symmetric distributions the PP index after centring-then-sphering is independent of the underlying population location and dispersion.     

The effects of nonnormality on asymptotic distributions of some likelihood ratio criteria for testing covariance structures under normal assumption

This paper examines asymptotic distributions of the likelihood ratio criteria, which are proposed under normality, for several hypotheses on covariance matrices when the true distribution of a population is a certain nonnormal distribution. It is well known that asymptotic distributions of test statistics depend on the fourth moments of the true population's distribution. We study the effects of nonnormality on the asymptotic distributions of the null and nonnull distributions of likelihood ratio criteria for covariance structures.     

Optimal Scaling of Random Walk Metropolis Algorithms with Non-Gaussian Proposals

The asymptotic optimal scaling of random walk Metropolis (RWM) algorithms with Gaussian proposal distributions is well understood for certain specific classes of target distributions. These asymptotic results easily extend to any light tailed proposal distribution with finite fourth moment. However, heavy tailed proposal distributions such as the Cauchy distribution are known to have a number of desirable properties, and in many situations are preferable to light tailed proposal distributions. Therefore we consider the question of scaling for Cauchy distributed proposals for a wide range of independent and identically distributed (iid) product densities. The results are somewhat surprising as to when and when not Cauchy distributed proposals are preferable to Gaussian proposal distributions. This provides motivation for finding proposal distributions which improve on both Gaussian and Cauchy proposals, both for finite dimensional target distributions and asymptotically as the dimension of the target density, d → ∞. Throughout we seek the scaling of the proposal distribution which maximizes the expected squared jumping distance (ESJD).     

Near-exact distributions for the likelihood ratio test statistic to test equality of several variance-covariance matrices in elliptically contoured distributions

The exact distribution of the likelihood ratio test statistic to test the equality of several variance-covariance matrices has a non-manageable form. On the other hand, the existing asymptotic approximations do not exhibit the necessary precision for many applications. For these reasons, the development of near-exact approximations to the distribution of this statistic, arising from a different method of approximating distributions, emerges as a desirable goal. These distributions, while being manageable are much closer to the exact distribution than the usual asymptotic distributions and opposite to these, are also asymptotic for increasing number of variables and matrices involved. Computational modules to implement the near-exact distributions are made available on a web-site.     

Estimates of MM type for the multivariate linear model

We propose a class of robust estimates for multivariate linear models. Based on the approach of MM-estimation (Yohai 1987, [24]), we estimate the regression coefficients and the covariance matrix of the errors simultaneously. These estimates have both a high breakdown point and high asymptotic efficiency under Gaussian errors. We prove consistency and asymptotic normality assuming errors with an elliptical distribution. We describe an iterative algorithm for the numerical calculation of these estimates. The advantages of the proposed estimates over their competitors are demonstrated through both simulated and real data.     

Asymptotic Normality of Cross-correlogram Estimates of the Response Function

The problem of estimation of an unknown response function of a time-invariant continuous linear system is considered. Discrete-time sample input–output cross-correlograms are taken as estimates of the response function. The inputs are supposed to be zero-mean stationary Gaussian processes close, in some sense, to a white noise. Both asymptotic normality of finite-dimensional distributions of the estimates and their asymptotic normality in spaces of continuous functions are studied. Our basic tool is a new integral representation for cumulants of the estimate as a finite sum of integrals involving cyclic products of kernels. Some inequalities for these integrals are obtained and their asymptotic behaviour is studied.     

线性模型中M估计分布的随机加权方法逼近

在线性模型中,M估计的渐近分布通常都涉及到不易估计的未知误差分布的某些量,如果要估计渐近方差,就需对这些冗余参数进行估计.利用随机加权方法可以避免先对误差分布中的冗余参数进行估计.给出了当自变量是随机变量时,M估计分布的随机加权逼近,证明了M估计分布的随机加权逼近是一致相合的.当取不同的凸函数,样本大小和随机权时,进一步利用蒙特卡洛方法研究估计分布.研究表明随机权取泊松权时,不仅达到同样的效果而且可以减小计算量.