Authors:Qingguo Tang; R. J. Karunamuni Pages: 489 - 521 Abstract: Finite mixture regression (FMR) models are frequently used in statistical modeling, often with many covariates with low significance. Variable selection techniques can be employed to identify the covariates with little influence on the response. The problem of variable selection in FMR models is studied here. Penalized likelihood-based approaches are sensitive to data contamination, and their efficiency may be significantly reduced when the model is slightly misspecified. We propose a new robust variable selection procedure for FMR models. The proposed method is based on minimum-distance techniques, which seem to have some automatic robustness to model misspecification. We show that the proposed estimator has the variable selection consistency and oracle property. The finite-sample breakdown point of the estimator is established to demonstrate its robustness. We examine small-sample and robustness properties of the estimator using a Monte Carlo study. We also analyze a real data set. PubDate: 2018-06-01 DOI: 10.1007/s10463-017-0602-4 Issue No:Vol. 70, No. 3 (2018)

Authors:Chin-Tsang Chiang; Shao-Hsuan Wang; Ming-Yueh Huang Pages: 523 - 551 Abstract: One attractive advantage of the presented single-index hazards regression is that it can take into account possibly time-dependent covariates. In such a model formulation, the main theme of this research is to develop a theoretically valid and practically feasible estimation procedure for the index coefficients and the induced survival function. In particular, compared with the existing pseudo-likelihood approaches, our one proposes an automatic bandwidth selection and suppresses an influence of outliers. By making an effective use of the considered versatile survival process, we further reduce a substantial finite-sample bias in the Chambless-Diao type estimator of the most popular time-dependent accuracy summary. The asymptotic properties of estimators and data-driven bandwidths are also established under some suitable conditions. It is found in simulations that the proposed estimators and inference procedures exhibit quite satisfactory performances. Moreover, the general applicability of our methodology is illustrated by two empirical data. PubDate: 2018-06-01 DOI: 10.1007/s10463-017-0600-6 Issue No:Vol. 70, No. 3 (2018)

Authors:Weihua Zhao; Jianbo Li; Heng Lian Pages: 553 - 582 Abstract: We consider an estimating equations approach to parameter estimation in adaptive varying-coefficient linear quantile model. We propose estimating equations for the index vector of the model in which the unknown nonparametric functions are estimated by minimizing the check loss function, resulting in a profiled approach. The estimating equations have a bias-corrected form that makes undersmoothing of the nonparametric part unnecessary. The estimating equations approach makes it possible to obtain the estimates using a simple fixed-point algorithm. We establish asymptotic properties of the estimator using empirical process theory, with additional complication due to the nuisance nonparametric part. The finite sample performance of the new model is illustrated using simulation studies and a forest fire dataset. PubDate: 2018-06-01 DOI: 10.1007/s10463-017-0599-8 Issue No:Vol. 70, No. 3 (2018)

Authors:M. Razmkhah; S. Simriz Pages: 583 - 604 Abstract: Suppose that the failure times of the units placed on a life-testing experiment are independent but nonidentically distributed random variables. Under progressively type II censoring scheme, distributional properties of the proposed random variables are presented and some inferences are made. Assuming that the random variables come from a proportional hazard rate model, the formulas are simplified and also the amount of Fisher information about the common parameters of this family is calculated. The results are also extended to a fixed covariates model. The performance of the proposed procedure is investigated via a real data set. Some numerical computations are also presented to study the effect of the proportionality rates in view of the Fisher information criterion. Finally, some concluding remarks are stated. PubDate: 2018-06-01 DOI: 10.1007/s10463-017-0598-9 Issue No:Vol. 70, No. 3 (2018)

Authors:Mihai Giurcanu; Brett Presnell Pages: 605 - 630 Abstract: We study the standard-bootstrap, the centered-bootstrap, and the empirical-likelihood bootstrap tests of hypotheses used in conjunction with generalized method of moments inference in correctly specified and misspecified moment condition models. We show that, under correct specification, the standard-bootstrap estimator of the null distribution of the J-test converges in distribution to a random distribution, verifying its inconsistency, while the centered and the empirical-likelihood bootstrap estimators are consistent. We provide higher-order expansions of the size distortions of the analytic and the bootstrap tests. We show that the standard-bootstrap parameter-tests are consistent under misspecification, while the centered-bootstrap parameter-tests are inconsistent. We propose a general bootstrap methodology which is highly accurate under correct specification and consistent under misspecification. In a simulation study, we explore the finite sample behavior of the analytic and the bootstrap tests for a panel data model and we apply our methodology on a real-world data set. PubDate: 2018-06-01 DOI: 10.1007/s10463-017-0604-2 Issue No:Vol. 70, No. 3 (2018)

Authors:Xiangshun Kong; Mingyao Ai; Kwok Leung Tsui Pages: 631 - 646 Abstract: Sliced Latin hypercube designs are popularly adopted for computer experiments with qualitative factors. Previous constructions require the sizes of different slices to be identical. Here we construct sliced designs with flexible sizes of slices. Besides achieving desirable one-dimensional uniformity, flexible sliced designs (FSDs) constructed in this paper accommodate arbitrary sizes for different slices and cover ordinary sliced Latin hypercube designs as special cases. The sampling properties of FSDs are derived and a central limit theorem is established. It shows that any linear combination of the sample means from different models on slices follows an asymptotic normal distribution. Some simulations compare FSDs with other sliced designs in collective evaluations of multiple computer models. PubDate: 2018-06-01 DOI: 10.1007/s10463-017-0603-3 Issue No:Vol. 70, No. 3 (2018)

Authors:Konstantin Eckle; Nicolai Bissantz; Holger Dette; Katharina Proksch; Sabrina Einecke Pages: 647 - 689 Abstract: In this paper, we propose methods for inference of the geometric features of a multivariate density. Our approach uses multiscale tests for the monotonicity of the density at arbitrary points in arbitrary directions. In particular, a significance test for a mode at a specific point is constructed. Moreover, we develop multiscale methods for identifying regions of monotonicity and a general procedure for detecting the modes of a multivariate density. It is shown that the latter method localizes the modes with an effectively optimal rate. The theoretical results are illustrated by means of a simulation study and a data example. The new method is applied to and motivated by the determination and verification of the position of high-energy sources from X-ray observations by the Swift satellite which is important for a multiwavelength analysis of objects such as Active Galactic Nuclei. PubDate: 2018-06-01 DOI: 10.1007/s10463-017-0605-1 Issue No:Vol. 70, No. 3 (2018)

Authors:Buddhananda Banerjee; Satyaki Mazumder Pages: 691 - 715 Abstract: An existence of change point in a sequence of temporally ordered functional data demands more attention in its statistical analysis to make a better use of it. Introducing a dynamic estimator of covariance kernel, we propose a new methodology for testing an existence of change in the mean of temporally ordered functional data. Though a similar estimator is used for the covariance in finite dimension, we introduce it for the independent and weakly dependent functional data in this context for the first time. From this viewpoint, the proposed estimator of covariance kernel is more natural one when the sequence of functional data may possess a change point. We prove that the proposed test statistics are asymptotically pivotal under the null hypothesis and consistent under the alternative. It is shown that our testing procedures outperform the existing ones in terms of power and provide satisfactory results when applied to real data. PubDate: 2018-06-01 DOI: 10.1007/s10463-017-0606-0 Issue No:Vol. 70, No. 3 (2018)

Authors:Lijie Gu; Suojin Wang; Lijian Yang Abstract: Stratified sampling is one of the most important survey sampling approaches and is widely used in practice. In this paper, we consider the estimation of the distribution function of a finite population in stratified sampling by the empirical distribution function (EDF) and kernel distribution estimator (KDE), respectively. Under general conditions, the rescaled estimation error processes are shown to converge to a weighted sum of transformed Brownian bridges. Moreover, simultaneous confidence bands (SCBs) are constructed for the population distribution function based on EDF and KDE. Simulation experiments and illustrative data example show that the coverage frequencies of the proposed SCBs under the optimal and proportional allocations are close to the nominal confidence levels. PubDate: 2018-05-21 DOI: 10.1007/s10463-018-0668-7

Authors:Yoshiharu Takagi; Yutaka Kano Abstract: Recently, it is becoming more active to apply appropriate statistical methods dealing with missing data in clinical trials. Under not missing at random missingness, MLE based on direct-likelihood, or observed likelihood, possibly has a serious bias. A solution to the bias problem is to add auxiliary variables such as surrogate endpoints to the model for the purpose of reducing the bias. We theoretically studied the impact of an auxiliary variable on MLE and evaluated the bias reduction or inflation in the case of several typical correlation structures. PubDate: 2018-05-17 DOI: 10.1007/s10463-018-0667-8

Authors:Arun Kumar Kuchibhotla; Somabha Mukherjee; Ayanendranath Basu Abstract: M-estimators offer simple robust alternatives to the maximum likelihood estimator. The density power divergence (DPD) and the logarithmic density power divergence (LDPD) measures provide two classes of robust M-estimators which contain the MLE as a special case. In each of these families, the robustness of the estimator is achieved through a density power down-weighting of outlying observations. Even though the families have proved to be useful in robust inference, the relation and hierarchy between these two families are yet to be fully established. In this paper, we present a generalized family of divergences that provides a smooth bridge between DPD and LDPD measures. This family helps to clarify and settle several longstanding issues in the relation between the important families of DPD and LDPD, apart from being an important tool in different areas of statistical inference in its own right. PubDate: 2018-05-17 DOI: 10.1007/s10463-018-0665-x

Authors:Moosup Kim; Sangyeol Lee Abstract: This study considers the problem of testing whether the tail index of the GARCH innovations undergoes a change according to the values of conditional volatilities. Special attention is paid to power-transformed and threshold generalized autoregressive conditional heteroscedasticity processes that can accommodate the GARCH family. We show that the proposed test asymptotically follows a functional of a standard Brownian motion under some regularity conditions. To evaluate our method, we carry out a simulation study and real data analysis using the return series of the Google stock price and DowJones index. PubDate: 2018-05-17 DOI: 10.1007/s10463-018-0669-6

Authors:Yongcheng Qi; Fang Wang; Lin Zhang Abstract: Consider a p-variate normal random vector. We are interested in the limiting distributions of likelihood ratio test (LRT) statistics for testing the independence of its grouped components based on a random sample of size n. In classical multivariate analysis, the dimension p is fixed or relatively small, and the limiting distribution of the LRT is a chi-square distribution. When p goes to infinity, the chi-square approximation to the classical LRT statistic may be invalid. In this paper, we prove that the LRT statistic converges to a normal distribution under quite general conditions when p goes to infinity. We propose an adjusted test statistic which has a chi-square limit in general. Our comparison study indicates that the adjusted test statistic outperforms among the three approximations in terms of sizes. We also report some numerical results to compare the performance of our approaches and other methods in the literature. PubDate: 2018-05-14 DOI: 10.1007/s10463-018-0666-9

Authors:Kairat Mynbaev; Carlos Martins-Filho Abstract: Kernel density estimation in domains with boundaries is known to suffer from undesirable boundary effects. We show that in the case of smooth densities, a general and elegant approach is to estimate an extension of the density. The resulting estimators in domains with boundaries have biases and variances expressed in terms of density extensions and extension parameters. The result is that they have the same rates at boundary and interior points of the domain. Contrary to the extant literature, our estimators require no kernel modification near the boundary and kernels commonly used for estimation on the real line can be applied. Densities defined on the half-axis and in a unit interval are considered. The results are applied to estimation of densities that are discontinuous or have discontinuous derivatives, where they yield the same rates of convergence as for smooth densities on \({\mathbb {R}}\) . PubDate: 2018-05-04 DOI: 10.1007/s10463-018-0663-z

Authors:Lei Wang Abstract: To obtain M-estimators of a response variable when the data are missing at random, we can construct three bias-corrected nonparametric estimating equations based on inverse probability weighting, mean imputation, and augmented inverse probability weighting approaches. However, when the dimension of covariate is not low, the estimation efficiency will be affected due to the curse of dimensionality. To address this issue, we propose a two-stage estimation procedure by using the dimension-reduced kernel estimators in conjunction with bias-corrected estimating equations. We show that the resulting three kernel-assisted estimating equations yield asymptotically equivalent M-estimators that achieve the desirable properties. The finite-sample performance of the proposed estimators for response mean, distribution function and quantile is studied through simulation, and an application to HIV-CD4 data set is also presented. PubDate: 2018-04-25 DOI: 10.1007/s10463-018-0664-y

Authors:Alexey Miroshnikov; Evgeny Savelev Abstract: In this article, we perform an asymptotic analysis of Bayesian parallel kernel density estimators introduced by Neiswanger et al. (in: Proceedings of the thirtieth conference on uncertainty in artificial intelligence, AUAI Press, pp 623–632, 2014). We derive the asymptotic expansion of the mean integrated squared error for the full data posterior estimator and investigate the properties of asymptotically optimal bandwidth parameters. Our analysis demonstrates that partitioning data into subsets requires a non-trivial choice of bandwidth parameters that optimizes the estimation error. PubDate: 2018-04-18 DOI: 10.1007/s10463-018-0662-0

Authors:Stefan Aulbach; Michael Falk; Timo Fuller Abstract: A multivariate distribution function F is in the max-domain of attraction of an extreme value distribution if and only if this is true for the copula corresponding to F and its univariate margins. Aulbach et al. (Bernoulli 18(2), 455–475, 2012. https://doi.org/10.3150/10-BEJ343) have shown that a copula satisfies the extreme value condition if and only if the copula is tail equivalent to a generalized Pareto copula (GPC). In this paper, we propose a \(\chi ^2\) -goodness-of-fit test in arbitrary dimension for testing whether a copula is in a certain neighborhood of a GPC. The test can be applied to stochastic processes as well to check whether the corresponding copula process is close to a generalized Pareto process. Since the p value of the proposed test is highly sensitive to a proper selection of a certain threshold, we also present graphical tools that make the decision, whether or not to reject the hypothesis, more comfortable. PubDate: 2018-04-17 DOI: 10.1007/s10463-018-0657-x

Authors:Yuta Koike; Zhi Liu Abstract: The recent empirical works have pointed out that the realized skewness, which is the sample skewness of intraday high-frequency returns of a financial asset, serves as forecasting future returns in the cross section. Theoretically, the realized skewness is interpreted as the sample skewness of returns of a discretely observed semimartingale in a fixed interval. The aim of this paper is to investigate the asymptotic property of the realized skewness in such a framework. We also develop an estimation theory for the limiting characteristic of the realized skewness in a situation where measurement errors are present and sampling times are stochastic. PubDate: 2018-04-17 DOI: 10.1007/s10463-018-0659-8

Authors:Michael Levine Abstract: We consider the problem of adaptive estimation of the functional component in a partial linear model where the argument of the function is defined on a q-dimensional grid. Obtaining an adaptive estimator of this functional component is an important practical problem in econometrics where exact distributions of random errors and the parametric component are mostly unknown. An estimator of the functional component that is adaptive over the wide range of multivariate Besov classes and robust to a wide choice of distributions of the linear component and random errors is constructed. It is also shown that the same estimator is locally adaptive over the same range of Besov classes and robust over large collections of distributions of the linear component and random errors as well. At any fixed point, this estimator attains a local adaptive minimax rate. PubDate: 2018-04-13 DOI: 10.1007/s10463-018-0661-1

Authors:Nadezhda Gribkova; Ričardas Zitikis Abstract: Various members of the class of weighted insurance premiums and risk capital allocation rules have been researched from a number of perspectives. Corresponding formulas in the case of parametric families of distributions have been derived, and they have played a pivotal role when establishing parametric statistical inference in the area. Nonparametric inference results have also been derived in special cases such as the tail conditional expectation, distortion risk measure, and several members of the class of weighted premiums. For weighted allocation rules, however, nonparametric inference results have not yet been adequately developed. In the present paper, therefore, we put forward empirical estimators for the weighted allocation rules and establish their consistency and asymptotic normality under practically sound conditions. Intricate statistical considerations rely on the theory of induced order statistics, known as concomitants. PubDate: 2018-04-07 DOI: 10.1007/s10463-018-0660-2