Change-Point Tests for the Error Distribution in Non-parametric Regression

Abstract.  Several testing procedures are proposed that can detect change-points in the error distribution of non-parametric regression models. Different settings are considered where the change-point either occurs at some time point or at some value of the covariate. Fixed as well as random covariates are considered. Weak convergence of the suggested difference of sequential empirical processes based on non-parametrically estimated residuals to a Gaussian process is proved under the null hypothesis of no change-point. In the case of testing for a change in the error distribution that occurs with increasing time in a model with random covariates the test statistic is asymptotically distribution free and the asymptotic quantiles can be used for the test. This special test statistic can also detect a change in the regression function. In all other cases the asymptotic distribution depends on unknown features of the data-generating process and a bootstrap procedure is proposed in these cases. The small sample performances of the proposed tests are investigated by means of a simulation study and the tests are applied to a data example.     

Comparison of Separable Components in Different Samples

Abstract.  Imagine we have two different samples and are interested in doing semi- or non-parametric regression analysis in each of them, possibly on the same model. In this paper, we consider the problem of testing whether a specific covariate has different impacts on the regression curve in these two samples. We compare the regression curves of different samples but are interested in specific differences instead of testing for equality of the whole regression function. Our procedure does allow for random designs, different sample sizes, different variance functions, different sets of regressors with different impact functions, etc. As we use the marginal integration approach, this method can be applied to any strong, weak or latent separable model as well as to additive interaction models to compare the lower dimensional separable components between the different samples. Thus, in the case of having separable models, our procedure includes the possibility of comparing the whole regression curves, thereby avoiding the curse of dimensionality. It is shown that bootstrap fails in theory and practice. Therefore, we propose a subsampling procedure with automatic choice of subsample size. We present a complete asymptotic theory and an extensive simulation study.     

Testing Monotonicity of Regression Functions – An Empirical Process Approach

We propose several new tests for monotonicity of regression functions based on different empirical processes of residuals and pseudo‐residuals. The residuals are obtained from an unconstrained kernel regression estimator whereas the pseudo‐residuals are obtained from an increasing regression estimator. Here, in particular, we consider a recently developed simple kernel‐based estimator for increasing regression functions based on increasing rearrangements of unconstrained non‐parametric estimators. The test statistics are estimated distance measures between the regression function and its increasing rearrangement. We discuss the asymptotic distributions, consistency and small sample performances of the tests.     

Estimating the Conditional Error Distribution in Non‐parametric Regression

Abstract. We consider a general non‐parametric regression model, where the distribution of the error, given the covariate, is modelled by a conditional distribution function. For the estimation, a kernel approach as well as the (kernel based) empirical likelihood method are discussed. The latter method allows for incorporation of additional information on the error distribution into the estimation. We show weak convergence of the corresponding empirical processes to Gaussian processes and compare both approaches in asymptotic theory and by means of a simulation study.     

Non-parametric Analysis of Covariance – The Case of Inhomogeneous and Heteroscedastic Noise

Abstract.  The purpose of this paper was to propose a procedure for testing the equality of several regression curves f _i in non-parametric regression models when the noise is inhomogeneous and heteroscedastic, i.e. when the variances depend on the regressor and may vary between groups. The presented approach is very natural because it transfers the maximum likelihood statistic from a heteroscedastic one-way analysis of variance to the context of non-parametric regression. The maximum likelihood estimators will be replaced by kernel estimators of the regression functions f _i . It is shown that the asymptotic distribution of the obtained test-statistic is nuisance parameter free. Asymptotic efficiency is compared with a test of Dette & Neumeyer [Annals of Statistics (2001) Vol. 29, 1361–1400] and it is shown that the new test is asymptotically uniformly more powerful. For practical purposes, a bootstrap variant is suggested. In a simulation study, level and power of this test will be briefly investigated and compared with other procedures. In summary, our theoretical findings are supported by this study. Finally, a crop yield experiment is reanalysed.     

Smooth Residual Bootstrap for Empirical Processes of Non-parametric Regression Residuals

Abstract.  The aim of this paper is to prove the validity of smooth residual bootstrap versions of procedures that are based on the empirical process of residuals estimated from a non-parametric regression model. From this result, consistency of various model tests in non-parametric regression is deduced, such as goodness-of-fit tests for the regression and variance function, tests for equality of regression functions and tests concerning the error distribution.     

Sequential fertility decision making: Catholics and Protestants in Canada*

Cette étude tente d'apprécier l'importance relative de l'utilisation, pour l'analyse de la fécondité d'un cadre séquential, plutôt que du mode d'approche statique utilisé par “l‘école” de Becker. On s'est basé sur les données du recensement canadien de 1971, pour analyser les naissances par parité des femmes protestantes et catholiques. Les résultats montrent des différences appréciables dans l'effet des variables indépendantes, suivant la parité, ce qui confirme fortement l'intérêt du mode d'approche dynamique. Plus précisément, on note que le revenu du mariàun effet positif pour les parités de rang faible, alors qu'on trouve un effet négatif pour les parités de rang élevé. Ces résultats contredisent l'effet constamment positif que postule la théorie micro-économique. L'en-semble des effets est le même pour les échantillons de protestantes et de catholiques. This paper assesses the relative importance of utilizing a sequential framework in fertility analysis, rather than the static framework employed by the Becker “school.” Parity-specific analyses are conducted for Protestant and Catholic women based on 1971 Canadian census data. The findings indicate that there are considerable differences in the effects of the independent variables on fertility by parity, lending strong support to a dynamic approach. Specifically, husband's income has a positive effect at earlier parities reflecting timing differences, whereas a negative effect emerges at higher parities. These findings are contrary to the consistently positive effect postulated by microeconomic theory. The pattern of effects is similar in the Protestant and Catholic samples.