Statistical Methodology最新文献

Recently, a new family of skew distributions was proposed using a specific class of transformation of scale, in which the normalizing constant remains unchanged and unimodality is readily assured. In this paper, we introduce the mode invariance in this family, which allows us to easily study certain properties, including monotonicity of skewness, and incorporate various favorable properties. The entropy maximization for a skew distribution is discussed. A numerical study is also conducted.

Di Crescenzo and Longobardi (2002) introduced the past entropy, Sunoj et al. (2013) gave a quantile version for the past entropy, termed as the past quantile entropy (PQE). Based on the PQE function, they defined a new stochastic order called as less PQE (LPQE) order and studied some properties of this order. In the present paper, we focus our interests on further closure properties of this new order. Some characterizations of the LPQE order are investigated, closure and reversed closure properties are obtained. The preservation of the LPQE order in the proportional failure rate and reversed failure rate models is discussed.

In this paper, we develop simple non-parametric test based on U-statistics for testing constant failure rate against IFR, IFRA, DMRL, NBU and NBUE alternatives. The asymptotic properties of the test statistics are studied. In particular, the test statistics are shown to be asymptotically normal and consistent against the relevant alternatives. Some numerical results are presented to demonstrate the performance of the proposed tests.

In this paper, we study the problem of testing for a copula parameter change in nonlinear autoregressive (AR) models with nonlinear generalized autoregressive conditional heteroskedasticity (GARCH) errors. To perform a test, we propose the cusum test based on pseudo maximum likelihood estimates of copula parameters. We derive its limiting null distribution under regularity conditions. For illustration, we conduct a simulation study with an emphasis on STAR–STGARCH models. A real data analysis applied to the S&P 500 index and IBM stock price is also considered.

Adaptive testing for the partially linear single-index model (PLSIM) with error-prone linear covariates is considered. This is a fundamentally important and interesting problem for the current model because existing literature often assumes that the model structure is known before making inferences. In practice, this may result in an incorrect inference on the PLSIM. In this study, we explore whether the link function satisfies some special shape constraints by using an efficient penalized estimating method. For this we propose a model structure selection method by constructing a new testing statistic in the current setting with measurement error, which may enhance the flexibility and predictive power of this model under the case that one can correctly choose an adaptive shape and model structure. The finite sample performance of the proposed methodology is investigated by using some simulation studies and a real example from the Framingham Heart Study.

In this article we construct bivariate discrete distributions in $Z^2$. We make use of a generalized trivariate reduction technique. The special case leading to a generalization of a bivariate Skellam distribution is studied in detail. Properties of the derived models as well as estimation are examined. Real data application is provided. Discussion of extensions to different models is also mentioned.

In the context of unsupervised classification of multidimensional data, we revisit the classical mixture model in the case where the dependencies among the random variables are described by a DAG structure. This structure is considered at two levels, the original DAG and its macro-representation. This two-level representation is the main base of the proposed mixture model. To perform unsupervised classification, we propose a dedicated algorithm called EM-mDAG, which extends the classical EM algorithm. In the Gaussian case, we show that this algorithm can be efficiently implemented. This approach has two main advantages. It favors the selection of a small number of classes and it allows a semantic interpretation of the classes based on a clustering within the macro-variables.

Justification of heavy tail is an important open problem. A systematic approach is proposed to verify heavy tail in linear time series. It consists of three parts, each of which is guided by statistical tests. The analysis is supplemented by an application to ozone concentration. The methodology has the advantage that the threshold selection is data-driven. Simulations show that test results are accurate even under model misspecification. The power is good under two heavy-tailed alternatives. The test is invariant when the time series clusters at extreme level in the study of the max-autoregressive process. It also gives a preliminary measure of tail heaviness if the underlying process is heavy-tailed.

Bernstein polynomials have many interesting properties. In statistics, they were mainly used to estimate density functions and regression relationships. The main objective of this paper is to promote further use of Bernstein polynomials in statistics. This includes (1) providing a high-level approximation of the moments of a continuous function $g\left(X\right)$ of a random variable $X$, and (2) proving Jensen’s inequality concerning a convex function without requiring second differentiability of the function. The approximation in (1) is demonstrated to be quite superior to the delta method, which is used to approximate the variance of $g\left(X\right)$ with the added assumption of differentiability of the function. Two numerical examples are given to illustrate the application of the proposed methodology in (1).

We show that if a linear combination of expectations of order statistics has mean zero across all random variables that have finite mean, then the linear combination is identically zero. A consequence of this result is that any functional of a probability distribution can have essentially only one unbiased $L$-estimator (i.e., an estimator that has the form of a linear combination of order statistics): if two such linear combinations have the same expectation then they must be algebraically identical. We use this result to prove the equivalence of two statistics that have been proposed as estimators of the L-moments introduced by Hosking (1990), and to provide alternative means of computing estimators of the trimmed L-moments introduced by Elamir and Seheult (2003). We also make comparisons of the speed of various methods for computing estimators of L-moments and trimmed L-moments.