Estimating Prediction Error: Cross-Validation vs. Accumulated Prediction Error

We study the validation of prediction rules such as regression models and classification algorithms through two out-of-sample strategies, cross-validation and accumulated prediction error. We use the framework of Efron (1983 Efron , B. ( 1983 ). Estimating the error rate of a prediction rule: improvement on cross-validation . Journal of the American Statistical Association 78 : 316 – 331 .[Taylor & Francis Online], [Web of Science ®] , [Google Scholar]) where measures of prediction errors are defined as sample averages of expected errors and show through exact finite sample calculations that cross-validation and accumulated prediction error yield different smoothing parameter choices in nonparametric regression. The difference in choice does not vanish as sample size increases.     

Sliced Average Variance Estimation for Censored Data

We consider a nonlinear censored regression problem with a vector of predictors. With censoring, high-dimensional regression analysis becomes much more complicated. Since censoring can cause severe bias in estimation, modification to adjust such bias is needed to be made. Based on the weight adjustment, we develop the modification of sliced average variance estimation for estimating the lifetime central subspace without requiring a prespecified parametric model. Our proposed method preserves as much regression information as possible. Simulation results are reported and comparisons are made with the sliced inverse regression of Li et al. (1999 Li , K. C. , Wang , J. L. , Chen , C. H. ( 1999 ). Dimension reduction for censored regression data . Ann. Statist. 27 : 1 – 23 . [Google Scholar]).     

Influence diagnostics in Birnbaum–Saunders nonlinear regression models

We consider the issue of assessing influence of observations in the class of Birnbaum–Saunders nonlinear regression models, which is useful in lifetime data analysis. Our results generalize those in Galea et al. [8 Galea, M., Leiva, V. and Paula, G. A. 2004. Influence diagnostics in log-Birnbaum–Saunders regression models. J. Appl. Stat., 31: 1049–1064. [Taylor & Francis Online], [Web of Science ®] [Google Scholar]] which are confined to Birnbaum–Saunders linear regression models. Some influence methods, such as the local influence, total local influence of an individual and generalized leverage are discussed. Additionally, the normal curvatures for studying local influence are derived under some perturbation schemes. We also give an application to a real fatigue data set.     

On Improvement in Variance Estimation Using Auxiliary Information

Kadilar and Cingi (2006 Kadilar , C. , Cingi , H. ( 2006 ). Improvement in variance estimation using auxiliary information . Hacett. J. Math. Statist. 35 ( 1 ): 111 – 115 . [Google Scholar]) have introduced an estimator for the population variance using an auxiliary variable in simple random sampling. We propose a new ratio-type exponential estimator for population variance which is always more efficient than usual ratio and regression estimators suggested by Isaki (1983 Isaki , C. T. ( 1983 ). Variance estimation using auxiliary information . J. Amer. Statist. Assoc. 78 : 117 – 123 .[Taylor & Francis Online], [Web of Science ®] , [Google Scholar]) and by Kadilar and Cingi (2006 Kadilar , C. , Cingi , H. ( 2006 ). Improvement in variance estimation using auxiliary information . Hacett. J. Math. Statist. 35 ( 1 ): 111 – 115 . [Google Scholar]). Efficiency comparison is carried out both mathematically and numerically.     

Bias-Corrected Maximum Likelihood Estimators in Nonlinear Heteroscedastic Models

Nonlinear heteroscedastic models are widely used in econometrics and statistical applications. We derive matrix formulae for the second-order biases of the maximum likelihood estimators of the parameters in the mean and variance response which generalize previous results by Cook et al. (1986 Cook , D. R. , Tsai , C. L. , Wei , B. C. ( 1986 ). Bias in nonlinear regression . Biometrika 73 : 615 – 623 .[Crossref], [Web of Science ®] , [Google Scholar]) and Cordeiro (1993 Cordeiro , G. M. ( 1993 ). Bartlett corrections and bias correction for two heteroscedastic regression models . Commun. Statist. Theor. Meth. 22 : 169 – 188 .[Taylor & Francis Online], [Web of Science ®] , [Google Scholar]). The biases of the estimators are easily obtained as vectors of regression coefficients from suitable weighted linear regressions. The practical use of such biases is illustrated in a simulation study and in an application to a real data set.     

Cross-validation: What is it and how is it used in regression?

The purpose of this article is to explain cross-validation and describe its use in regression. Because replicability analyses are not typically employed in studies, this is a topic with which many researchers may not be familiar. As a result, researchers may not understand how to conduct cross-validation in order to evaluate the replicability of their data. This article not only explains the purpose of cross-validation, but also uses the widely available Holzinger and Swineford (1939 Holzinger, K.J., Swineford, F. (1939). A Study in Factor Analysis: The Stability of a Bi-Factor Solution. Chicago, IL: University of Chicago. Available at: http://people.cehd.tamu.edu/～bthompson/datasets.htm [Google Scholar]) dataset as a heuristic example to concretely demonstrate its use. By incorporating multiple tables and examples of SPSS syntax and output, the reader is provided with additional visual examples in order to further clarify the steps involved in conducting cross-validation. A brief discussion of the limitations of cross-validation is also included. After reading this article, the reader should have a clear understanding of cross-validation, including when it is appropriate to use, and how it can be used to evaluate replicability in regression.     

MSE Performance of a Heterogeneous Pre-Test Ridge Regression Estimator

This article describes how diagnostic procedures were derived for symmetrical nonlinear regression models, continuing the work carried out by Cysneiros and Vanegas (2008 Cysneiros , F. J. A. , Vanegas , L. H. ( 2008 ). Residuals and their statistical properties in symmetrical nonlinear models . Statist. Probab. Lett. 78 : 3269 – 3273 .[Crossref], [Web of Science ®] , [Google Scholar]) and Vanegas and Cysneiros (2010 Vanegas , L. H. , Cysneiros , F. J. A. ( 2010 ). Assesment of diagnostic procedures in symmetrical nonlinear regression models . Computat. Statist. Data Anal. 54 : 1002 – 1016 .[Crossref], [Web of Science ®] , [Google Scholar]), who showed that the parameters estimates in nonlinear models are more robust with heavy-tailed than with normal errors. In this article, we focus on assessing if the robustness of this kind of models is also observed in the inference process (i.e., partial F-test). Symmetrical nonlinear regression models includes all symmetric continuous distributions for errors covering both light- and heavy-tailed distributions such as Student-t, logistic-I and -II, power exponential, generalized Student-t, generalized logistic, and contaminated normal. Firstly, a statistical test is shown to evaluating the assumption that the error terms all have equal variance. The results of simulation studies which describe the behavior of the test for heteroscedasticity proposed in the presence of outliers are then given. To assess the robustness of inference process, we present the results of a simulation study which described the behavior of partial F-test in the presence of outliers. Also, some diagnostic procedures are derived to identify influential observations on the partial F-test. As ilustration, a dataset described in Venables and Ripley (2002 Venables , W. N. , Ripley , B. D. ( 2002 ). Modern Applied with S. , 4th ed. New York : Springer .[Crossref] , [Google Scholar]), is also analyzed.     

Covariate-Adjusted Partially Linear Regression Models

Motivated by covariate-adjusted regression (CAR) proposed by Sentürk and Müller (2005 Sentürk , D. , Müller , H. G. ( 2005 ). Covariate-adjusted regression . Biometrika 92 : 75 – 89 .[Crossref], [Web of Science ®] , [Google Scholar]) and an application problem, in this article we introduce and investigate a covariate-adjusted partially linear regression model (CAPLM), in which both response and predictor vector can only be observed after being distorted by some multiplicative factors, and an additional variable such as age or period is taken into account. Although our model seems to be a special case of covariate-adjusted varying coefficient model (CAVCM) given by Sentürk (2006 Sentürk , D. ( 2006 ). Covariate-adjusted varying coefficient models . Biostatistics 7 : 235 – 251 .[Crossref], [PubMed], [Web of Science ®] , [Google Scholar]), the data types of CAPLM and CAVCM are basically different and then the methods for inferring the two models are different. In this article, the estimate method motivated by Cui et al. (2008 Cui , X. , Guo , W. S. , Lin , L. , Zhu , L. X. ( 2008 ). Covariate-adjusted nonlinear regression . Ann. Statist. 37 : 1839 – 1870 . [Google Scholar]) is employed to infer the new model. Furthermore, under some mild conditions, the asymptotic normality of estimator for the parametric component is obtained. Combined with the consistent estimate of asymptotic covariance, we obtain confidence intervals for the regression coefficients. Also, some simulations and a real data analysis are made to illustrate the new model and methods.     

New Prediction Interval and Band in the Nonlinear Regression Model: Application to Predictive Modeling in Foods

This article is concerned with the proposal of a new prediction interval and band for the nonlinear regression model. The construction principle of this interval and band is based on an exact (the meaning of the term “exact” will be given later) confidence region for parameters of the nonlinear regression model. This region, fully described in Vila and Gauchi (2007 Vila , J.-P. , Gauchi , J.-P. ( 2007 ). Optimal designs based on exact confidence regions for parameter estimation of a nonlinear regression model . Journal of Statistical Planning and Inference 137 ( 9 ): 2935 – 2953 .[Crossref], [Web of Science ®] , [Google Scholar]), provides a rigorous justification for the new prediction interval and band that we propose. This new band is then compared to the classical bands (which are asymptotic and thus approximate for small n), and also to the band based on the bootstrap resampling method. The comparison of these bands is undertaken with simulated and real data from predictive modeling in food science.     

On Universal Optimality of Circular Weakly Neighbor Balanced Designs Under an Interference Model

In this article, the optimal design problem in a fixed effects interference model with left-neighbor effects is studied. It is known (Druilhet, 1999 Druilhet , P. ( 1999 ). Optimality of circular neighbor balanced designs . J. Statist. Plann. Infer. 81 : 141 – 152 .[Crossref], [Web of Science ®] , [Google Scholar]) that circular neighbor balanced designs (CNBDs) are universally optimal in such a model. We prove the universal optimality of circular weakly neighbor balanced designs (CWNBDs), which require a smaller number of blocks than CNBDs. CWNBDs with the number of blocks smaller than the number of treatments belong to the class of partially neighbor balanced designs (PNBDs) defined by Wilkinson et al. (1983 Wilkinson , G. N. , Eckert , R. , Hancock , T. W. , Mayo , O. ( 1983 ). Nearest neighbour (NN) analysis of field experiments . J. Roy. Statist. Soc. Ser. B 45 : 151 – 210 . [Google Scholar]). We give a construction method for some CWNBDs, with examples.     

Accelerated Life Test Data Analysis for Repairable Systems

In this article, the imperfect maintenance model and proportional intensity model are used to analyze failure data of repairable systems in accelerated life testing. In the design and development phase of products, we should collect and analyze failure data quickly with small proto-type products. Thus, we test the products under accelerated conditions and if the products fail, then we repair and use those continuously in the life testing. Acceleration and repair models are needed to analyze the failure data. An age reduction model (Brown et al.'s Brown et al. 1983 Brown , J. F. , Mahoney , J. F. , Sivazlian , B. D. ( 1983 ). Hysteresis repair in discounted replacement problems . IIE Trans. 15 : 156 – 165 . [CSA] [Taylor & Francis Online], [Web of Science ®] , [Google Scholar], model) and relationship between scale parameter and stress level are assumed. The stress acts multiplicatively on the baseline cumulative intensity. The maximum likelihood method is used, the log-likelihood function is derived, and a maximizing procedure is proposed. In simulation studies, we investigate the accuracy and trends of the maximum likelihood estimator.     

Statistical Properties of Threshold Models

This article focuses the attention on the Self Exciting Threshold Autoregressive Moving Average model (SETARMA) proposed in Tong (1983 Tong , H. ( 1983 ). Threshold Models in Nonlinear Time Series Analysis . London : Springer-Verlag .[Crossref] , [Google Scholar]). The stochastic structure of the model is discussed and different specifications are presented. Starting from one of them, we give sufficient conditions for the weak stationarity of the model that are discussed and critically compared to other results given in literature. In particular, after showing that the SETARMA model belongs to the class of the Random Coefficients Autoregressive models, widely discussed in Nicholls and Quinn (1982 Nicholls , D. F. , Quinn , B. G. (1982). Random Coefficients Autoregressive Models. An Introduction . New York : Springer-Verlag.[Crossref] , [Google Scholar]), we give some issues on the weak stationarity of its stochastic structure that are more general than those given in the existing literature and appear not affected by the moving average component.     

Improved Estimation of Finite Population Variance Using Auxiliary Information

This article addresses the problem of estimating of finite population variance using auxiliary information in simple random sampling. A ratio-cum-difference type class of estimators for population variance has been suggested with its properties under large sample approximation. It has been shown that the suggested class of estimators is more efficient than usual unbiased, difference, Das and Tripathi (1978 Das , A. K. , Tripathi , T. P. ( 1978 ). Use of auxiliary information in estimating the finite population variance . Sankhya C 40 : 139 – 148 . [Google Scholar]), Isaki (1983 Isaki , C. T. ( 1983 ). Variance estimation using auxiliary information . J. Amer. Statist. Assoc. 78 : 117 – 123 .[Taylor & Francis Online], [Web of Science ®] , [Google Scholar]), Singh et al. (1988 Singh , H. P. , Upadhyaya , L. N. , Namjoshi , U. D. ( 1988 ). Estimation of finite population variance . Curr. Sci. 57 : 1331 – 1334 .[Web of Science ®] , [Google Scholar]), Kadilar and Cingi (2006 Kadilar , C. , Cingi , H. ( 2006 ). Ratio estimators for the population variance in simple and stratified random sampling . Appl. Math. Comp. 173 ( 2 ): 1047 – 1059 .[Crossref], [Web of Science ®] , [Google Scholar]), and other estimators/classes of estimators. In addition, we support this theoretical result with the aid of a empirical study.     

On the Functional Approach to Bayesian Efficient Designs for Nonlinear Regression Models

The present article is devoted to an extension of the functional approach elaborated in the book Melas (2006 Melas , V. B. ( 2006 ). Functional Approach to Optimal Experimental Design . Lecture Notes in Statistics , Vol. 184. Heidelberg : Springer . [Google Scholar]) for studying optimal designs in linear and nonlinear regression models. Here we consider Bayesian efficient designs for nonlinear models under the standard assumptions on the observational errors. Sufficient conditions for uniqueness of locally optimal and Bayesian efficient designs for common optimality criteria are given. L-efficient Bayesian designs are constructed and investigated for a special nonlinear regression model of a rational form as an illustration of our main results. This model is interesting in both a practical and a theoretical sense.     

On Asymptotic Normality of the Local Polynomial Regression Estimator with Stochastic Bandwidths

Nonparametric density and regression estimators commonly depend on a bandwidth. The asymptotic properties of these estimators have been widely studied when bandwidths are non stochastic. In practice, however, in order to improve finite sample performance of these estimators, bandwidths are selected by data driven methods, such as cross-validation or plug-in procedures. As a result, nonparametric estimators are usually constructed using stochastic bandwidths. In this article, we establish the asymptotic equivalence in probability of local polynomial regression estimators under stochastic and nonstochastic bandwidths. Our result extends previous work by Boente and Fraiman (1995 Boente , G. , Fraiman , R. ( 1995 ). Asymptotic distribution of data-driven smoothers in density and regression estimation under dependence . Can. J. Statist. 23 : 383 – 397 .[Crossref], [Web of Science ®] , [Google Scholar]) and Ziegler (2004 Ziegler , K. ( 2004 ). Adaptive kernel estimation of the mode in nonparametric random design regression model . Probab. Mathemat. Statist. 24 : 213 – 235 . [Google Scholar]).     

Bayes minimax ridge regression estimators

The problem of estimating of the vector β of the linear regression model y = Aβ + ? with ? ～ N_p(0, σ²I_p) under quadratic loss function is considered when common variance σ² is unknown. We first find a class of minimax estimators for this problem which extends a class given by Maruyama and Strawderman (2005 Maruyama, Y., and W. E. Strawderman. 2005. A new class of generalized Bayes minimax ridge regression estimators. Annals of Statistics 33:1753–70.[Crossref], [Web of Science ®] , [Google Scholar]) and using these estimators, we obtain a large class of (proper and generalized) Bayes minimax estimators and show that the result of Maruyama and Strawderman (2005 Maruyama, Y., and W. E. Strawderman. 2005. A new class of generalized Bayes minimax ridge regression estimators. Annals of Statistics 33:1753–70.[Crossref], [Web of Science ®] , [Google Scholar]) is a special case of our result. We also show that under certain conditions, these generalized Bayes minimax estimators have greater numerical stability (i.e., smaller condition number) than the least-squares estimator.     

Empirical Likelihood Method for Censored Median Regression Models

Based on the semiparametric median regression analysis for the right-censored data developed by Ying et al. (1995 Ying , Z. , Jung , S. H. , Wei , L. J. ( 1995 ). Survival analysis with median regression models . J. Amer. Statist. Assoc. 90 : 178 – 184 .[Taylor & Francis Online], [Web of Science ®] , [Google Scholar]), an empirical likelihood based inferential procedure for the regression coefficients is proposed. The limiting distribution of the proposed log-empirical likelihood ratio test statistic follows a chi-squared distribution, which corresponds to the standard asymptotic results of the empirical likelihood method. The inference about the subsets of the entire regression coefficients vector is discussed. The proposed method is illustrated by some simulation studies.     

Nonnested Testing for Competing Autoregressive Dynamic Models Estimated by Instrumental Variables

In this work we investigate nonnested tests for two competing univariate dynamic linear models with autoregressive disturbances, where the motivation for instrumental variable estimation is mainly due to the recognized presence of current endogenous variables in the regression function, either in one or both models. As the previous transformation of both models yields regression functions which are nonlinear in the parameters, the attractive Gauss-Newton regression (GNR) approach, firstly advocated by Davidson and Mackinnon (1981 Davidson , R. , Mackinnon , J. G. ( 1981 ). Several tests for model specification in the presence of alternative hypotheses . Econometrica 49 : 781 – 93 .[Crossref], [Web of Science ®] , [Google Scholar]), will be used to obtain the results.     

A Computer Program to Improve LR Tests for Generalized Linear Models

Generalized linear models enable the fitting of models to a wide range of data types. These models are based on exponential dispersion distributions. Improved likelihood ratio tests for these models were developed by Cordeiro (1983 Cordeiro , G. M. (1983). Improved likelihood ratio statistics for generalized linear models. Journal of the Royal Statistical Society, Series B: Methodological 45:404–413. [Google Scholar])Cordeiro (1987 Cordeiro , G. M. ( 1987 ). On the corrections to the likelihood ratio statistics . Biometrika 74 : 265 – 274 .[Crossref], [Web of Science ®] , [Google Scholar]). We present a simple R program source for calculating Bartlett corrections to improve likelihood ratio tests in these models. The program was tested on some special models, confirming all of the previously reported numerical results for the Bartlett corrections.     

An Index of Local Sensitivity to Nonignorability for a Pseudolikelihood Method

For regression analysis of data with non response, sensitivity analysis is usually recommended. An index of local sensitivity to non ignorability (ISNI) (Troxel et al., 2004 Troxel , A. B. , Ma , G. , Heitjan , D. F. (2004). An index of local sensitivity to nonignorability. Statistica Sinica 14:1221–1237.[Web of Science ®] , [Google Scholar]) was derived to detect the sensitivity of maximum likelihood estimates to small departures from ignorability. However, ISNI requires specification of a parametric model for the missing-data mechanism. In this article, a local sensitivity index for a pseudolikelihood (PL) method that does not require specification of the mechanism is proposed. For bivariate data (x, y), when the non response mechanism is an arbitrary function of x + λy, this new index is defined as the first derivative of the PL estimate with respect to λ at λ = 0. The closed form was derived for normal regression data when the density function of the predictor x approximated by a kernel estimator in the PL method. The utility of this new local sensitivity index was illustrated through application on one dataset.