Admissible variable-selection procedures when fitting misspecified regression models by least squares

For loss equal to squared error of prediction, Kempthorne(l984) has proved that all variable-selection procedures are admissible for choosing among least-squares fits of a normal linear regression model. We extend this result to the case of a normal linear regression model in which the form of the expected response vector is misspecified.     

A two-stage method for the estimation of global sensitivity indices of non-parametric models

This article aims to propose a method to effectively estimate global sensitivity indices under non-parametric models. The new method involves two stages. First, all the non-influential sensitivity indices are filtered out by an adjustive W-statistic test process with low cost, and then the remaining significant sensitivity indices are precisely estimated by an orthogonal array (OA) with large number of levels and low strength. The method avoids complicated prototype building and shows a much lower experimental cost. The performance of this method as well as comparisons with polynomial regression method, Gaussian Process (GP) method, and component selection and smoothing operator (COSSO) method are tested on three numerical models that are widely used in engineering and statistical areas. Finally, a real data example is analyzed.     

Asymmetric selection indices

In 1936, H. Fairfield Smith (A discriminant function for plant selection, Annalsof Eugenice (London) 7,240–260) suggested a linear selection index for selecting varieties with higher genotypic values. Since then, the idea has been extended in various directions such as restricted selection indices. In this paper, linear selection indices are considered when, unlike squared error, the loss function is asymmetric. In particular, a LINEX loss function is considered for this purpose. It is shown that under multivariate normality, this approach still leads to the usual selection indices. Certain computational aspects are indicated.     

The Mean Squared Error of the Instrumental Variables Estimator When the Disturbance Has an Elliptical Distribution

This paper generalizes Nagar's (1959) approximation to the finite sample mean squared error (MSE) of the instrumental variables (IV) estimator to the case in which the errors possess an elliptical distribution whose moments exist up to infinite order. This allows for types of excess kurtosis exhibited by some financial data series. This approximation is compared numerically to Knight's (1985) formulae for the exact moments of the IV estimator under nonnormality. We use the results to explore two questions on instrument selection. First, we complement Buse's (1992) analysis by considering the impact of additional instruments on both bias and MSE. Second, we evaluate the properties of Andrews's (1999) selection method in terms of the bias and MSE of the resulting IV estimator.     

Simultaneous testing of parameter subsets in less than full rank models

If the experimental design (or lack of design) results in a nodel which is not of f u l l rank, the problem of variable selection becomes rather complex. Tkis is due to the fact that, in less than f u l l rank models, not every linear combination of regression parameters is estimable. In this paper, we present a procedure fortesting all “testable” subsets of a complete set of regression parameters, using a technique based on Scheffe's method (1959). A class of “adequate” subsets of regression parameters is obtained in a manner similar to that of Aitkin (1974). The proposed procedure is illustraced with an example.     

Lower Bounds on Integrated Risk,Subject to Inequality Constraints

Statistical decision theory can sometimes be used to find, via a least favourable prior distribution, a statistical procedure that attains the minimax risk. This theory also provides, using an ‘unfavourable prior distribution’, a very useful lower bound on the minimax risk. In the late 1980s, Kempthorne showed how, using a least favourable prior distribution, a specified integrated risk can sometimes be minimised, subject to an inequality constraint on a different risk. Specifically, he was concerned with the solution of a minimax‐Bayes compromise problem (‘compromise decision theory’). Using an unfavourable prior distribution, Kabaila & Tuck ( 2008 ), provided a very useful lower bound on an integrated risk, subject to an inequality constraint on a different risk. We extend this result to the case of multiple inequality constraints on specified risk functions and integrated risks. We also describe a new and very effective method for the computation of an unfavourable prior distribution that leads to a very useful lower bound. This method is simply to maximize the lower bound directly with respect to the unfavourable prior distribution. Not only does this method result in a relatively tight lower bound, it is also fast because it avoids the repeated computation of the global maximum of a function with multiple local maxima. The advantages of this computational method are illustrated using the problems of bounding the performance of a point estimator of (i) the multivariate normal mean and (ii) the univariate normal mean.     

Generalized Begun–Reid estimators for multiple hazard rates

A class of invariant estimators with respect to the selection of a base population is developed for estimating the hazard rates in multiple populations. The class generalizes the estimators of Begun and Reid (J. Amer. Statist. Assoc. 78 (1983) 337) and includes the estimator of Mantel and Haenszel (J. Natl. Canser Inst. 22 (1959) 719) as a special case. The estimators have explicit forms and, it is shown that their asymptotic covariance matrices are less than those of the Begun–Reid estimators when the number of populations is greater than two. A Monte-Carlo simulation indicates that the estimators are slightly more efficient than the Cox partial likelihood estimator (Biometrika 62 (2) (1975) 269) for small and medium sample sizes. An example is presented for the illustration of the estimators.     

An efficient method for the estimation of multivariate moving averge models

Durbin's (1959) efficient method for the estimation of univariate moving average models is generalized to the vector case. Strong consistency and asymptotic normality of the estimator is proved. A simulation experiment is performed to illustrate the behaviour of the method in finite samples.     

On a conjecture concerning the least favorable configuration of a two-stage selection procedure

A two-stage procedure 𝓅with screening in the first stage to find the population with the largest mean out of k ≧ 2 normal populations with unknown means and a common variance is under concern. It was proposed and previousiy studied by Cohen (1959), Alam (1970) and Tamhane and Bechhofer (1977, 1979) using the indifference-zone approach. The conjecture that the least favourable parameter configuration for the probability of a correct selection is of the slippage type remained unproved for k ≧ 3. Miescke and Sehr (1980) proved the conjecture for k=3. The problem was further discussed by Gupta and Miescke (1982). A general proof for rhe conjecture will be given in this paper.     

系统性风险度量CoES的建模和检验

本文结合极值理论(Extreme Value Theory,EVT)和新的动态混合Copula(Dynamic Mixture Copula,DM-Copula)函数,提出了一种新的CoES估计方法DM-Copula-EVT。在EVT建模中,本文改进了阈值的选取方法以避免选择的主观性,并提出了一系列新的动态混合Copula以更好地刻画金融市场日益复杂的尾部关联性。此外,本文首次提出了检验CoES模型设定正确性的后验分析方法,包括无条件覆盖性检验和条件覆盖性检验。将本文建模和检验方法应用于我国金融市场,研究发现:相对于传统使用的t分布,EVT能更好地拟合指数的尾部分布;新的动态混合Copula函数能更好地刻画金融部门与系统之间的复杂关联性。     

New improvements in the use of dependence measures for sensitivity analysis and screening

ABSTRACT

Physical phenomena are commonly modelled by time consuming numerical simulators, function of many uncertain parameters whose influences can be measured via a global sensitivity analysis. The usual variance-based indices require too many simulations, especially as the inputs are numerous. To address this limitation, we consider recent advances in dependence measures, focusing on the distance correlation and the Hilbert–Schmidt independence criterion. We study and use these indices for a screening purpose. Numerical tests reveal differences between variance-based indices and dependence measures. Then, two approaches are proposed to use the latter for a screening purpose. The first approach uses independence tests, with existing asymptotic versions and spectral extensions; bootstrap versions are also proposed. The second considers a linear model with dependence measures, coupled to a bootstrap selection method or a Lasso penalization. Numerical experiments show their potential in the presence of many non-influential inputs and give successful results for a nuclear reliability application.     

Spacing of information in polynomial regression:a simple solution

Consider a k polynomial regression on a single real variable. If n uncorrelated observations are to be taken in a design with support on more than k+1 points, there is an approximate experiment, ν, with support on k+1 points and n observations such that both designs have the same information matrix for the model. A proof of this result is provided. A method to obtain the approximate design ν is given and illustrated by an example. The source of disagreement between Kiefer (1959) and De La Garza (1954) in the solution of this problem is clarified.     

Testing for co-directional interactions using union–intersection and intersection–union methods

When interaction terms exist in a two-factor, factorial experiment, the consideration and analysis of main effects are often restricted to those situations where the interaction between factors is not significant. Hinkelman and Kempthorne [4 Hinkelmann, K. and Kempthorne, O. 1994. Design and Analysis of Experiments. Volume 1: Introduction to Experimental Design, New York: Wiley.  [Google Scholar]] softened that stance somewhat and advocate testing main effects when the interaction is deemed co-directional but not anti-directional. A test for the main effects in that situation may be pragmatic to the practitioner and appealing to researchers in other disciplines. Intersection–union and union–intersection methods are examined for assessing the directional nature of significant interactions so that the main effects in a two-factor factorial may be evaluated. The tests suggested are conceptually straightforward and practical and maintain the nominal Type-I error rate. Examples are provided to illustrate the methods.     

Seemingly unrelated regression tree

Nonparametric seemingly unrelated regression provides a powerful alternative to parametric seemingly unrelated regression for relaxing the linearity assumption. The existing methods are limited, particularly with sharp changes in the relationship between the predictor variables and the corresponding response variable. We propose a new nonparametric method for seemingly unrelated regression, which adopts a tree-structured regression framework, has satisfiable prediction accuracy and interpretability, no restriction on the inclusion of categorical variables, and is less vulnerable to the curse of dimensionality. Moreover, an important feature is constructing a unified tree-structured model for multivariate data, even though the predictor variables corresponding to the response variable are entirely different. This unified model can offer revelatory insights such as underlying economic meaning. We propose the key factors of tree-structured regression, which are an impurity function detecting complex nonlinear relationships between the predictor variables and the response variable, split rule selection with negligible selection bias, and tree size determination solving underfitting and overfitting problems. We demonstrate our proposed method using simulated data and illustrate it using data from the Korea stock exchange sector indices.     

A wavelet-based approach to the analysis and modelling of financial time series exhibiting strong long-range dependence: the case of Southeast Europe

This paper demonstrates the utilization of wavelet-based tools for the analysis and prediction of financial time series exhibiting strong long-range dependence (LRD). Commonly emerging markets' stock returns are characterized by LRD. Therefore, we track the LRD evolvement for the return series of six Southeast European stock indices through the application of a wavelet-based semi-parametric method. We further engage the á trous wavelet transform in order to extract deeper knowledge on the returns term structure and utilize it for prediction purposes. In particular, a multiscale autoregressive (MAR) model is fitted and its out-of-sample forecast performance is benchmarked to that of ARMA. Additionally, a data-driven MAR feature selection procedure is outlined. We find that the wavelet-based method captures adequately LRD dynamics both in calm as well as in turmoil periods detecting the presence of transitional changes. At the same time, the MAR model handles with the complicated autocorrelation structure implied by the LRD in a parsimonious way achieving better performance.     

Bounds for the t-tail area

The bounds of Birnbaum (1942) and Sampford (1953) for the upper tail area of the normal distribution are extended to the upper tail of the t-distribution. Numerical and theoretical comparisons are made with the bounds of Peizer and Pratt (1968), Wallace (1959) and Soms (1977).     

Mathematical forms of the distribution of the product of two normal variables

This paper finds the mathematical forms of the distribution of the product where x and x follow a bivariate normal distribution In this paper the distribution when PT0 is expressed as an integral, a new, fundamental result. From this general form, six different cases can be distinguished depending on what is known about the parameters and p. The special cases are Aroian $year:1959 and (6) Additionally, we prove that if and as the distribution of the product approaches the Type III distribution. When p=0# Aroian $year:1959 and Aroian and Meeker $year:1977, give tables for various values of 6., 6 . The results in this paper will be used to provide brief tables for p^O in a separate paper     

Detecting active effects in unreplicated designs

Unreplicated factorial designs pose a difficult problem in analysis because there are no degrees of freedom left to estimate the error. Daniel [Technometrics 1 (1959), pp. 311-341] proposed an ingenious graphical method that does not require σ to be estimated. Here we try to put Daniel's method into a formal framework and lift the subjectiveness that carries. A simulation study has been conducted that shows that the proposed method behaves better than Lenth's [Technometrics 31 (1989), pp. 469-473] popular method.     

Quasi-Likelihood Regression with Multiple Indices and Smooth Link and Variance Functions

Abstract.  A flexible semi-parametric regression model is proposed for modelling the relationship between a response and multivariate predictor variables. The proposed multiple-index model includes smooth unknown link and variance functions that are estimated non-parametrically. Data-adaptive methods for automatic smoothing parameter selection and for the choice of the number of indices M are considered. This model adapts to complex data structures and provides efficient adaptive estimation through the variance function component in the sense that the asymptotic distribution is the same as if the non-parametric components are known. We develop iterative estimation schemes, which include a constrained projection method for the case where the regression parameter vectors are mutually orthogonal. The proposed methods are illustrated with the analysis of data from a growth bioassay and a reproduction experiment with medflies. Asymptotic properties of the estimated model components are also obtained.     

General Multi-Level Modeling with Sampling Weights

ABSTRACT

In this article we study the approximately unbiased multi-level pseudo maximum likelihood (MPML) estimation method for general multi-level modeling with sampling weights. We conduct a simulation study to determine the effect various factors have on the estimation method. The factors we included in this study are scaling method, size of clusters, invariance of selection, informativeness of selection, intraclass correlation, and variability of standardized weights. The scaling method is an indicator of how the weights are normalized on each level. The invariance of the selection is an indicator of whether or not the same selection mechanism is applied across clusters. The informativeness of the selection is an indicator of how biased the selection is. We summarize our findings and recommend a multi-stage procedure based on the MPML method that can be used in practical applications.