Revisiting tests for neglected nonlinearity using artificial neural networks

Tests for regression neglected nonlinearity based on artificial neural networks (ANNs) have so far been studied by separately analyzing the two ways in which the null of regression linearity can hold. This implies that the asymptotic behavior of general ANN-based tests for neglected nonlinearity is still an open question. Here we analyze a convenient ANN-based quasi-likelihood ratio statistic for testing neglected nonlinearity, paying careful attention to both components of the null. We derive the asymptotic null distribution under each component separately and analyze their interaction. Somewhat remarkably, it turns out that the previously known asymptotic null distribution for the type 1 case still applies, but under somewhat stronger conditions than previously recognized. We present Monte Carlo experiments corroborating our theoretical results and showing that standard methods can yield misleading inference when our new, stronger regularity conditions are violated.     

Bootstrap hypothesis testing for some common statistical problems: A critical evaluation of size and power properties

The construction of bootstrap hypothesis tests can differ from that of bootstrap confidence intervals because of the need to generate the bootstrap distribution of test statistics under a specific null hypothesis. Similarly, bootstrap power calculations rely on resampling being carried out under specific alternatives. We describe and develop null and alternative resampling schemes for common scenarios, constructing bootstrap tests for the correlation coefficient, variance, and regression/ANOVA models. Bootstrap power calculations for these scenarios are described. In some cases, null-resampling bootstrap tests are equivalent to tests based on appropriately constructed bootstrap confidence intervals. In other cases, particularly those for which simple percentile-method bootstrap intervals are in routine use such as the correlation coefficient, null-resampling tests differ from interval-based tests. We critically assess the performance of bootstrap tests, examining size and power properties of the tests numerically using both real and simulated data. Where they differ from tests based on bootstrap confidence intervals, null-resampling tests have reasonable size properties, outperforming tests based on bootstrapping without regard to the null hypothesis. The bootstrap tests also have reasonable power properties.     

Testing for jumps in the EGARCH process

This paper considers testing for jumps in the exponential GARCH (EGARCH) models with Gaussian and Student-t innovations. The Wald and log likelihood ratio tests contain a nuisance parameter unidentified under the null hypothesis of no jumps, and hence are unavailable for this problem, because jump probability and variance of jumps in the test statistic cannot be estimated under the null hypothesis of no jumps. It is shown that the nuisance parameter is cancelled out in the Lagrange multiplier (LM) test statistic, and hence that the test is nuisance parameter-free. The one-sided test is also proposed using the nonnegative constraint on jump variance. The actual size and power of the tests are examined in a Monte Carlo experiment. The test is applied to daily returns of S&P 500 as an illustrative example.     

On testing equality of intraclass correlations under unequal family sizes

The three likelihood-based tests, namely, likelihood ratio test, Rao score test, and Wald test and two more asymptotic tests which use Srivastava's estimator of intraclass correlation coefficient are considered to test the null hypothesis of equality of intraclass correlation coefficients when the families have unequal number of children. Methods are illustrated on Galton's data set. Using simulation experiment we compute the sizes and powers of these tests and compare. It is found that our proposed test using Srivastava's estimator and the score test perform the best among all tests.     

Nonparametric linear tests with multiple events

A generalization of the nonparametric linear rank statistics is presented to handle the two-group comparison with multiple events. For a sample divided into two groups, in which each subject may experience at least two distinct failures, the logrank tests are extended to test the null hypothesis that the vector of the marginal survival distributions of the first group equals that of the second group. Two cases are distinguished depending on whether the null hypothesis does or does not imply the equality of the joint survival functions. In both cases, under the null hypothesis, the asymptotic joint distribution of the vector of the marginal statistics is shown to be Gaussian with covariance matrix consistently estimated using martingale properties. These theoretical results are illustrated by a simulation study and an application on the German Breast Cancer data. An extension to multiple hypotheses testing in multivariate proportional hazards models is also developed.     

Semiparametrically weighted robust estimation of regression models

A class of two-step robust regression estimators that achieve a high relative efficiency for data from light-tailed, heavy-tailed, and contaminated distributions irrespective of the sample size is proposed and studied. In particular, the least weighted squares (LWS) estimator is combined with data-adaptive weights, which are determined from the empirical distribution or quantile functions of regression residuals obtained from an initial robust fit. Just like many existing two-step robust methods, the LWS estimator with the proposed weights preserves robust properties of the initial robust estimate. However, contrary to the existing methods and despite the data-dependent weights, the first-order asymptotic behavior of LWS is fully independent of the initial estimate under mild conditions. Moreover, the proposed estimation method is asymptotically efficient if errors are normally distributed. A simulation study documents these theoretical properties in finite samples; in particular, the relative efficiency of LWS with the proposed weighting schemes can reach 85%-100% in samples of several tens of observations under various distributional models.     

Size and power properties of some tests in the Birnbaum-Saunders regression model

The Birnbaum-Saunders distribution has been used quite effectively to model times to failure for materials subject to fatigue and for modeling lifetime data. In this paper we obtain asymptotic expansions, up to order n^−1/2 and under a sequence of Pitman alternatives, for the non-null distribution functions of the likelihood ratio, Wald, score and gradient test statistics in the Birnbaum-Saunders regression model. The asymptotic distributions of all four statistics are obtained for testing a subset of regression parameters and for testing the shape parameter. Monte Carlo simulation is presented in order to compare the finite-sample performance of these tests. We also present two empirical applications.     

Testing fuzzy hypotheses based on fuzzy test statistic

A new approach for testing fuzzy parametric hypotheses based on fuzzy test statistic is introduced. First, we define some models representing the extended versions of the simple, the one-sided and the two-sided crisp hypotheses to the fuzzy ones. Then, we provide a confidence interval for interested parameter, and using α-cuts of the fuzzy null hypothesis, we construct the related fuzzy test statistic. Finally, by introducing a credit level, we can decide to accept or reject the fuzzy hypothesis. The method is applied to test the fuzzy hypotheses for the mean of a normal distribution, the variance of a normal distribution, and the mean of a Poisson distribution.     

Profile inference on partially linear varying-coefficient errors-in-variables models under restricted condition

In this paper, we investigate the estimation and testing problems of partially linear varying-coefficient errors-in-variables (EV) models under additional restricted condition. The restricted estimators of parametric and nonparametric components are established based on modified profile least-squares method, and their asymptotic properties are also studied under some regularity conditions. Moreover, the modified profile Lagrange multiplier test statistic is constructed under additional restricted condition. It is shown that the modified profile Lagrange multiplier test statistic is asymptotically distribution-free and follows a Chi-squared distribution under the null hypothesis. Some simulation studies are carried out to assess the performance of the proposed methods. A real dataset is analyzed for illustration.     

On testing for serial correlation of unknown form using wavelet thresholding

Omnibus procedures for testing serial correlation are developed, using spectral density estimation and wavelet shrinkage. The asymptotic distributions of the wavelet coefficients under the null hypothesis of no serial correlation are derived. Under some general conditions on the wavelet basis, the wavelet coefficients asymptotically follow a normal distribution. Furthermore, they are asymptotically uncorrelated. Adopting a spectral approach and using results on wavelet shrinkage, new one-sided test statistics are proposed. As a spatially adaptive estimation method, wavelets can effectively detect fine features in the spectral density, such as sharp peaks and high frequency alternations. Using an appropriate thresholding parameter, shrinkage rules are applied to the empirical wavelet coefficients, resulting in a non-linear wavelet-based spectral density estimator. Consequently, the advocated approach avoids the need to select the finest scale J, since the noise in the wavelet coefficients is naturally suppressed. Simple data-dependent threshold parameters are also considered. In general, the convergence of the spectral test statistics toward their respective asymptotic distributions appears to be relatively slow. In view of that, Monte Carlo methods are investigated. In a small simulation study, several spectral test statistics are compared, with respect to level and power, including versions of these test statistics using Monte Carlo simulations.     

Testing and estimation of purely nonparametric effects in repeated measures designs

The several sample case of the so-called nonparametric Behrens-Fisher problem in repeated measures designs is considered. That is, even under the null hypothesis, the marginal distribution functions in the different groups may have different shapes, and are not assumed to be equal. Moreover, the continuity of the marginal distribution functions is not required so that data with ties and, particularly, ordered categorical data are covered by this model. A multiple relative treatment effect is defined which can be estimated by using the mid-ranks of the observations within pairwise samples. The asymptotic distribution of this estimator is derived, along with a consistent estimator of its asymptotic covariance matrix. In addition, a multiple contrast test and related simultaneous confidence intervals for the relative marginal effects are derived and compared to rank-based Wald-type and ANOVA-type statistics. Simulations show that the ANOVA-type statistic and the multiple contrast test appear to maintain the pre-assigned level of the test quite accurately (even for rather small sample sizes) while the Wald-type statistic leads, as expected, to somewhat liberal decisions. Regarding the power, none of the statistics is uniformly superior. A real data set illustrates the application.     

Entropy test and residual empirical process for autoregressive conditional duration models

In this paper, we study the entropy test for the goodness of fit test in (nonlinear) autoregressive conditional duration (ACD) models. To implement a test, we first explore the null limiting distribution of the residual empirical process from ACD models and verify that it has an asymptotic expansion form that consists of the true empirical process and extra terms yielded by parameter estimation. Then, we show that under regularity conditions, the proposed entropy test approximately follows a distribution that is free from the parameter estimation. For illustration, a simulation study and real data analysis are conducted. In the implementation of the test, a parametric bootstrap method is employed.     

Alternative approaches to implementing Lagrange multiplier tests for serial correlation in dynamic regression models

An approximate F-form of the Lagrange multiplier (LM) test for serial correlation in dynamic regression models is compared with three bootstrap tests. In one bootstrap procedure, residuals from restricted estimation under the null hypothesis are resampled. The other two bootstrap tests use residuals from unrestricted estimation under an alternative hypothesis. A fixed autocorrelation alternative is assumed in one of the two unrestricted bootstrap tests and the other is based upon a Pitman-type sequence of local alternatives. Monte Carlo experiments are used to estimate rejection probabilities under the null hypothesis and in the presence of serial correlation.     

Shapiro-Wilk-type test of normality under nuisance regression and scale

An extension of the Shapiro-Wilk test to verify the hypothesis of normality in the presence of nuisance regression and scale has been previously considered. Such a test is typically based on the pair of the maximum likelihood and BLUE estimators of the standard deviation in the linear regression model. It has been shown that the asymptotic null distribution of the test criterion, extended to the regression model, is equivalent to that of the original Shapiro-Wilk test for the location-scale model. A simulation study is performed in order to show that both criteria are close under the normality hypothesis for moderate as well for large data sets. The power of the test against various alternative distributions of the model errors is illustrated. Furthermore, it is shown that the probabilities of errors of both the first and second kinds do not depend on the design matrix or on the parameters of the linear model.     

Subsampling tests for the mean change point with heavy-tailed innovations

The quest of the mean change point with innovations in the domain of attraction of a κ$\kappa$-stable law appears to still be ongoing. We adopt the residual CUSUM of squares test (RCUSQ) and derive its null asymptotic distribution, which is dependent on stable index κ$\kappa$. Then a residual-based subsampling is proposed to approximate the null distribution when stable index κ$\kappa$ is unknown. Consistency and the rate of convergence for the estimated change point are also obtained. We establish the asymptotic validity of this method and assess its performance both theoretically and numerically.     

平滑转移向量误差修正模型的非线性调节bootstrap检验

对原假设为线性不含平滑转移均衡趋势关系的两类平滑转移向量误差修正模型提出了非线性调节检验;提出了检验这类非线性调节关系的SupWald检验、相应的渐近分布和残差bootstrap方法模拟p值. 模拟实验中证实了有限样本下SupWald检验统计量的良好效用, 并给出了其适用范围, 进而应用该方法检验出几组美国国库券收益率间存在明显的平滑转移非线性调节.     

On wavelet-based testing for serial correlation of unknown form using Fan’s adaptive Neyman method

Test procedures for serial correlation of unknown form with wavelet methods are investigated. A new test statistic is motivated using a canonical multivariate normal hypothesis testing model. It relies on empirical wavelet coefficients of a wavelet-based spectral density estimator. The choice of the Haar wavelet function is advocated, since evidence demonstrates that the choice of the wavelet function is not critical. Under the null hypothesis of no serial correlation, the asymptotic distribution of a vector of empirical wavelet coefficients is derived, which is asymptotically a multivariate normal distribution. A test statistic is proposed based on that asymptotic result, which presents the serious advantage to be completely data-driven or adaptive, avoiding the selection of any smoothing parameters. Furthermore, under a suitable class of fixed alternatives, the wavelet-based method is consistent against serial correlation of unknown form. The test statistic is expected to exhibit good power properties when the true spectral density displays significant spatial inhomogeneity, such as seasonal or business cycle periodicities. However, the convergence of the test statistic towards its asymptotic distribution is relatively slow. Thus, Monte Carlo methods based on random samples are suggested to determine the corresponding critical values. In a simulation study, the new methodology is compared with several test statistics, with respect to their exact levels and powers. The robustness properties of the spectral methods based on Monte Carlo critical values are also investigated empirically, when the error terms are weak white noises.     

检验门限协整模型中的线性协整

考虑门限协整回归模型中线性的检验问题.在原假设为线性协整的条件下,构造TSupLM(supremumLagrange multiplier)统计量,并给出了极限分布.Monte Carlo实验研究了SupLM检验的有限样本性能,结果表明SupLM检验不受回归误差的序列相关性影响,也不受广义的自回归条件异方差GARCH(generalized autoregressiveconditional heteroskedastic)的影响.应用SupLM检验方法检测美国国库券收益率之间的关系,结果表明不同到期时间的国库券收益率之间存在门限协整关系.     

An Exact Probability Metric for Decision Tree Splitting and Stopping

ID3's information gain heuristic is well-known to be biased towards multi-valued attributes. This bias is only partially compensated for by C4.5's gain ratio. Several alternatives have been proposed and are examined here (distance, orthogonality, a Beta function, and two chi-squared tests). All of these metrics are biased towards splits with smaller branches, where low-entropy splits are likely to occur by chance. Both classical and Bayesian statistics lead to the multiple hypergeometric distribution as the exact posterior probability of the null hypothesis that the class distribution is independent of the split. Both gain and the chi-squared tests arise in asymptotic approximations to the hypergeometric, with similar criteria for their admissibility. Previous failures of pre-pruning are traced in large part to coupling these biased approximations with one another or with arbitrary thresholds; problems which are overcome by the hypergeometric. The choice of split-selection metric typically has little effect on accuracy, but can profoundly affect complexity and the effectiveness and efficiency of pruning. Empirical results show that hypergeometric pre-pruning should be done in most cases, as trees pruned in this way are simpler and more efficient, and typically no less accurate than unpruned or post-pruned trees.     

Size and power of tests for a zero random effect variance or polynomial regression in additive and linear mixed models

Several tests for a zero random effect variance in linear mixed models are compared. This testing problem is non-regular because the tested parameter is on the boundary of the parameter space. Size and power of the different tests are investigated in an extensive simulation study that covers a variety of important settings. These include testing for polynomial regression versus a general smooth alternative using penalized splines. Among the test procedures considered, three are based on the restricted likelihood ratio test statistic (RLRT), while six are different extensions of the linear model F-test to the linear mixed model. Four of the tests with unknown null distributions are based on a parametric bootstrap, the other tests rely on approximate or asymptotic distributions. The parametric bootstrap-based tests all have a similar performance. Tests based on approximate F-distributions are usually the least powerful among the tests under consideration. The chi-square mixture approximation for the RLRT is confirmed to be conservative, with corresponding loss in power. A recently developed approximation to the distribution of the RLRT is identified as a rapid, powerful and reliable alternative to computationally intensive parametric bootstrap procedures. This novel method extends the exact distribution available for models with one random effect to models with several random effects.