Bootstrap Version Research Articles

Cox-type tests for competing spatial autoregressive models with spatial autoregressive disturbances

In this paper, we consider the Cox-type tests of non-nested hypotheses for spatial autoregressive (SAR) models with SAR disturbances. We formally derive the asymptotic distributions of the test statistics. In contrast to regression models, we show that the Cox-type and J-type tests for non-nested hypotheses in the framework of SAR models are not asymptotically equivalent under the null hypothesis. The Cox test in a non-spatial setting has been found often to have large size distortion, which can be removed by bootstrap. Cox-type tests for SAR models with SAR disturbances may also have a large size distortion. We show that the bootstrap is consistent for Cox-type tests in our framework. Performances of the Cox-type and J-type tests as well as their bootstrapped versions in finite samples are compared via a Monte Carlo study. These tests are of particular interest when there are competing models with different spatial weight matrices. Using bootstrapped p-values, the Cox tests have relatively high power in all experiments and can outperform J-type and several other related tests in some cases.

Local Linear GMM Estimation of Functional Coefficient IV Models With an Application to Estimating the Rate of Return to Schooling

We consider the local linear generalized method of moment (GMM) estimation of functional coefficient models with a mix of discrete and continuous data and in the presence of endogenous regressors. We establish the asymptotic normality of the estimator and derive the optimal instrumental variable that minimizes the asymptotic variance-covariance matrix among the class of all local linear GMM estimators. Data-dependent bandwidth sequences are also allowed for. We propose a nonparametric test for the constancy of the functional coefficients, study its asymptotic properties under the null hypothesis as well as a sequence of local alternatives and global alternatives, and propose a bootstrap version for it. Simulations are conducted to evaluate both the estimator and test. Applications to the 1985 Australian Longitudinal Survey data indicate a clear rejection of the null hypothesis of the constant rate of return to education, and that the returns to education obtained in earlier studies tend to be overestimated for all the work experience.

Performance and estimation of the true error rate of classification rules built with additional information. An application to a cancer trial

Classification rules that incorporate additional information usually present in discrimination problems are receiving certain attention during the last years as they perform better than the usual rules. Fernández, M. A., C. Rueda and B. Salvador (2006): "Incorporating additional information to normal linear discriminant rules," J. Am. Stat. Assoc., 101, 569-577, proved that these rules have lower total misclassification probability than the usual Fisher's rule. In this paper we consider two issues; on the one hand, we compare these rules with those based on shrinkage estimators of the mean proposed by Tong, T., L. Chen and H. Zhao (2012): "Improved mean estimation and its application to diagonal discriminant analysis," Bioinformatics, 28(4): 531-537. with regard to four criteria: total misclassification probability, area under ROC curve, well-calibratedness and refinement; on the other hand, we consider the estimation of the true error rate, which is a very interesting parameter in applications. We prove results on the apparent error rate of the rules that expose the need of new estimators of their true error rate. We propose four such new estimators. Two of them are defined incorporating the additional information into the leave-one-out-bootstrap. The other two are the corresponding cross-validation after bootstrap versions. We compare these estimators with the usual ones in a simulation study and in a cancer trial application, showing the good behavior of the rules that incorporate additional information and of the new leave-one-out bootstrap estimators of their true error rate.

Significance testing in quantile regression

We consider the problem of testing significance of predictors in multivariate nonparametric quantile regression. A stochastic process is proposed, which is based on a comparison of the responses with a nonparametric quantile regression estimate under the null hypothesis. It is demonstrated that under the null hypothesis this process converges weakly to a centered Gaussian process and the asymptotic properties of the test under fixed and local alternatives are also discussed. In particular we show, that - in contrast to the nonparametric approach based on estimation of $L^{2}$-distances - the new test is able to detect local alternatives which converge to the null hypothesis with any rate $a_{n}\to 0$ such that $a_{n}\sqrt{n}\to\infty$ (here $n$ denotes the sample size). We also present a small simulation study illustrating the finite sample properties of a bootstrap version of the corresponding Kolmogorov-Smirnov test.

Model checks for the volatility under microstructure noise

We consider the problem of testing the parametric form of the volatility for high frequency data. It is demonstrated that in the presence of microstructure noise commonly used tests do not keep the preassigned level and are inconsistent. The concept of preaveraging is used to construct new tests, which do not suffer from these drawbacks. These tests are based on a Kolmogorov-Smirnov or Cramer-von-Mises functional of an integrated stochastic process, for which weak convergence to a (conditional) Gaussian process is established. The finite sample properties of a bootstrap version of the test are illustrated by means of a simulation study.

Testing Stationarity for Unobserved Components Models

In the aftermath of the global financial crisis, competing measures of the trend in macroeconomic variables such as US real GDP have featured prominently in policy debates. A key question is whether the large shocks to macroeconomic variables will have permanent effects — i.e., in econometric terms, do the data contain stochastic trends? Unobserved components models provide a convenient way to estimate stochastic trends for time series data, with their existence typically motivated by stationarity tests that allow for at most a deterministic trend under the null hypothesis. However, given the small sample sizes available for most macroeconomic variables, standard Lagrange multiplier tests of stationarity will perform poorly when the data are highly persistent. To address this problem, we propose the use of a likelihood ratio test of stationarity based directly on the unobserved components models used in estimation of stochastic trends. We demonstrate that a bootstrap version of this test has far better small-sample properties for empirically-relevant data generating processes than bootstrap versions of the standard Lagrange multiplier tests. An application to US real GDP produces stronger support for the presence of large permanent shocks when using the likelihood ratio test as compared to the standard tests.

On testing inference in beta regressions

This article deals with testing inference in the class of beta regression models with varying dispersion. We focus on inference in small samples. We perform a numerical analysis in order to evaluate the sizes and powers of different tests. We consider the likelihood ratio test, two adjusted likelihood ratio tests proposed by Ferrari and Pinheiro [Improved likelihood inference in beta regression, J. Stat. Comput. Simul. 81 (2011), pp. 431–443], the score test, the Wald test and bootstrap versions of the likelihood ratio, score and Wald tests. We perform tests on the parameters that index the mean submodel and also on the parameters in the linear predictor of the precision submodel. Overall, the numerical evidence favours the bootstrap tests. It is also shown that the score test is considerably less size-distorted than the likelihood ratio and Wald tests. An application that uses real (not simulated) data is presented and discussed.

Nonparametric bootstrap tests of conditional independence in two-way contingency tables

When analyzing a two-way contingency table, a preliminary question is often whether the categorical variables under study, say R and S, are independent or not. Suppose now that for each individual in the table, a continuous variable X is also known. It is then worth analyzing the table conditionally on X. Contrasting these “local” results to the global unconditional case allows one to go beyond the initial analysis and provide a better understanding of the underlying phenomenon. Recently, Geenens and Simar (2010) [11] have proposed two nonparametric procedures for testing whether R and S are conditionally independent given X, free of any constraining linearity assumptions. However, based on an average of kernel-based estimators, the asymptotic criterion they suggested shows an inflated Type I error (false positive) for small to moderate sample sizes. In this paper, we address this problem by proposing consistent bootstrap versions of the Geenens–Simar test procedures when testing for local independence. A comprehensive simulation study indeed shows the superiority of the bootstrap rejection criterion as compared to the asymptotic criterion in terms of Type I error. It also highlights the advantage of the flexibility guaranteed by the nonparametric Geenens–Simar tests when compared with parametric competitors, e.g. logistic models. The approach is finally illustrated with a real-data example.

Multivariate drought characteristics using trivariate Gaussian and Student t copulas

AbstractThis study aims to investigate the changing properties of drought events in Weihe River basin, China, by modeling the multivariate joint distribution of drought duration, severity and peak using trivariate Gaussian and Student t copulas. Monthly precipitations of Xi'an gauge are used to illustrate the meta‐elliptical copula‐based methodology for a single‐station application. Gaussian and Student t copulas are found to produce a better fit comparing with other six symmetrical and asymmetrical Archimedean copulas, and, checked by the goodness‐of‐fit tests based on a modified bootstrap version of Rosenblatt's transformation, both of them are acceptable to model the multivariate joint distribution of drought variables. Gaussian copula, the best fitting, is employed to construct the dependence structures of positively associated drought variables so as to obtain the multivariate joint and conditional probabilities of droughts. A Kendall's return period (KRP) introduced by Salvadori and De Michele (2010) is then adopted to assess the multivariate recurrent properties of drought events, and its spatial distributions indicate that prolonged droughts are likely to break out with rather short recurrence intervals in the whole region, while drought status in the southeast seems to be severer than the northwest. The study is of some merits in terms of multivariate drought modeling using a preferable copula‐based method, the results of which could serve as a reference for regional drought defense and water resources management. Copyright © 2011 John Wiley & Sons, Ltd.

Risk of incident cardiovascular events in patients with rheumatoid arthritis: a meta-analysis of observational studies

ObjectiveTo determine the magnitude of the risk of incident cardiovascular disease (CVD; fatal and non-fatal), including acute myocardial infarction (MI), cerebrovascular accidents (CVA) and congestive heart failure (CHF), in patients...

Analysis of multiple exposures in the case‐crossover design via sparse conditional likelihood

We adapt the least absolute shrinkage and selection operator (lasso) and other sparse methods (elastic net and bootstrapped versions of lasso) to the conditional logistic regression model and provide a full R implementation. These variable selection procedures are applied in the context of case-crossover studies. We study the performances of conventional and sparse modelling strategies by simulations, then empirically compare results of these methods on the analysis of the association between exposure to medicinal drugs and the risk of causing an injurious road traffic crash in elderly drivers. Controlling the false discovery rate of lasso-type methods is still problematic, but this problem is also present in conventional methods. The sparse methods have the ability to provide a global analysis of dependencies, and we conclude that some of the variants compared here are valuable tools in the context of case-crossover studies with a large number of variables.

Tests for Symmetric Error Distribution in Linear and Nonparametric Regression Models

In linear and nonparametric regression models, the problem of testing for symmetry of the distribution of errors is considered. We propose a test statistic which utilizes the empirical characteristic function of the corresponding residuals. The asymptotic null distribution of the test statistic as well as its behavior under alternatives is investigated. A simulation study compares bootstrap versions of the proposed test to other more standard procedures.

A direct bootstrapping technique and its application to a novel goodness of fit test

We prove general theorems that characterize situations in which we could have asymptotic closeness between the original statistics Hn and its bootstrap version Hn∗, without stipulating the existence of weak limits. As one possible application we introduce a novel goodness of fit test based on the modification of Total Variation metric. This new statistic is more sensitive than the Kolmogorov–Smirnov statistic, it applies to higher dimensions, and it does not converge weakly; but we show that it can be bootstrapped.

A simple test for regression specification with non-nested alternatives

In this paper, I introduce a simple test for the presence of the data-generating process among several non-nested alternatives. The test is an extension of the classical J test for non-nested regression models. I also provide a bootstrap version of the test that avoids possible size distortions inherited from the J test.

Testing for one-sided alternatives in nonparametric censored regression

Assume that we have two populations (X 1,Y 1) and (X 2,Y 2) satisfying two general nonparametric regression models Y j =m j (X j )+ε j , j=1,2, where m(⋅) is a smooth location function, ε j has zero location and the response Y j is possibly right-censored. In this paper, we propose to test the null hypothesis H 0:m 1=m 2 versus the one-sided alternative H 1:m 1<m 2. We introduce two test statistics for which we obtain the asymptotic normality under the null and the alternative hypotheses. Although the tests are based on nonparametric techniques, they can detect any local alternative converging to the null hypothesis at the parametric rate n −1/2. The practical performance of a bootstrap version of the tests is investigated in a simulation study. An application to a data set about unemployment duration times is also included.

Goodness-of-fit tests in semi-linear models

Specification tests for the error distribution are proposed in semi-linear models, including the partial linear model and additive models. The tests utilize an integrated distance involving the empirical characteristic function of properly estimated residuals. These residuals are obtained from an initial estimation step involving a combination of penalized least squares and smoothing techniques. A bootstrap version of the tests is utilized in order to study the small sample behavior of the procedures in comparison with more classical approaches. As an example, the tests are applied on some real data sets.

Levene type tests for the ratio of two scales

Tests for the equality of variances are of interest in many areas such as quality control, agricultural production systems, experimental education, pharmacology, biology, as well as a preliminary to the analysis of variance, dose–response modelling or discriminant analysis. The literature is vast. Traditional non-parametric tests are due to Mood, Miller and Ansari–Bradley. A test which usually stands out in terms of power and robustness against non-normality is the W50 Brown and Forsythe [Robust tests for the equality of variances, J. Am. Stat. Assoc. 69 (1974), pp. 364–367] modification of the Levene test [Robust tests for equality of variances, in Contributions to Probability and Statistics, I. Olkin, ed., Stanford University Press, Stanford, 1960, pp. 278–292]. This paper deals with the two-sample scale problem and in particular with Levene type tests. We consider 10 Levene type tests: the W50, the M50 and L50 tests [G. Pan, On a Levene type test for equality of two variances, J. Stat. Comput. Simul. 63 (1999), pp. 59–71], the R-test [R.G. O'Brien, A general ANOVA method for robust tests of additive models for variances, J. Am. Stat. Assoc. 74 (1979), pp. 877–880], as well as the bootstrap and permutation versions of the W50, L50 and R tests. We consider also the F-test, the modified Fligner and Killeen [Distribution-free two-sample tests for scale, J. Am. Stat. Assoc. 71 (1976), pp. 210–213] test, an adaptive test due to Hall and Padmanabhan [Adaptive inference for the two-sample scale problem, Technometrics 23 (1997), pp. 351–361] and the two tests due to Shoemaker [Tests for differences in dispersion based on quantiles, Am. Stat. 49(2) (1995), pp. 179–182; Interquantile tests for dispersion in skewed distributions, Commun. Stat. Simul. Comput. 28 (1999), pp. 189–205]. The aim is to identify the effective methods for detecting scale differences. Our study is different with respect to the other ones since it is focused on resampling versions of the Levene type tests, and many tests considered here have not ever been proposed and/or compared. The computationally simplest test found robust is W50. Higher power, while preserving robustness, is achieved by considering the resampling version of Levene type tests like the permutation R-test (recommended for normal- and light-tailed distributions) and the bootstrap L50 test (recommended for heavy-tailed and skewed distributions). Among non-Levene type tests, the best one is the adaptive test due to Hall and Padmanabhan.

An International Perspective on Risk Management Quality

AbstractThis paper introduces an alternative method for assessing the quality of risk management models. Specifically, using the forecast efficiency notion that forecast errors should be independent of a pertinent information set, we consider the extent to which unanticipated downside risk (extreme risk) is independent of overseas extreme risk. This is achieved using a bootstrap version of the non‐causality test recently introduced by Hong et al. (), data covering 45 international equity markets, and by measuring extreme risk via a class of risk management models recently introduced by Xiao and Koenker (). In doing this, we find significant evidence of transmission (causality) across national borders. Moreover, we discuss how risk managers in developed and emerging markets can parsimoniously incorporate such information (international dependency) into their risk management models to produce measures of downside risk that have more desirable ex post properties (viz. forecast efficiency properties).

Testing symmetry of a nonparametric bivariate regression function

We propose a test for symmetry of a regression function with a bivariate predictor based on the L 2 distance between the original function and its reflection. This distance is estimated by kernel methods and it is shown that under the null hypothesis as well as under the alternative the test statistic is asymptotically normally distributed. The finite sample properties of a bootstrap version of this test are investigated by means of a simulation study and a possible application in detecting asymmetries in grey-scale images is discussed.

Testing model assumptions in functional regression models

In the functional regression model where the responses are curves, new tests for the functional form of the regression and the variance function are proposed, which are based on a stochastic process estimating L 2 -distances. Our approach avoids the explicit estimation of the functional regression and it is shown that normalized versions of the proposed test statistics converge weakly. The finite sample properties of the tests are illustrated by means of a small simulation study. It is also demonstrated that for small samples, bootstrap versions of the tests improve the quality of the approximation of the nominal level.