Saddlepoint tests for quantile regression
نویسندگان
چکیده
Abstract: Quantile regression is a flexible and powerful technique which allows to model the quantiles of the conditional distribution of a response variable given a set of covariates. Regression quantile estimators can be viewed as M−estimators and standard asymptotic inference is readily available based on likelihoodratio, Wald, and score-type test statistics. However, these statistics require the estimation of the sparsity function s(α) = [g(G−1(α))]−1, where G and g are the cumulative distribution function and the density of the regression errors respectively, and this can lead to nonparametric density estimation. Moreover, the asymptotic χ distribution for these statistics can provide an inaccurate approximation of tail probabilities and this can lead to inaccurate p-values, especially for moderate sample sizes. Alternative methods which do not require the estimation of the sparsity function, include rank techniques and resampling methods to obtain confidence intervals, which can be inverted to test hypotheses. These are typically more accurate than the standard M− tests. In this paper we show how accurate tests can be obtained by using a nonparametric saddlepoint test statistic. The proposed statistic is asymptotically χ distributed, does not require the specification of the error distribution, and does not require the estimation of the sparsity function. The validity of the method is demonstrated through a simulation study, which shows both the robustness and the accuracy of the new test compared to the best available alternatives. The Canadian Journal of Statistics xx: 1–29; 20?? c © 20?? Statistical Society of Canada Résumé: La régression par quantiles est une technique souple et puissante pour modéliser les quantiles de la distribution conditionnelle d’une variable en fonction d’un ensemble de covariées. Les estimateurs de la régression par quantiles peuvent être écrits comme des M−estimateurs et l’inférence asymptotique standard est disponible. Elle est basée sur des statistiques de test du type du rapport de vraisemblance, Wald et multiplicateur de Lagrange. Toutefois le calcul de ces statistiques nécessite de l’estimation de la fonction de “sparsity” s(α) = [g(G−1(α))]−1, où G et g sont la fonction de répartition et la fonction de densité des erreurs respectivement et ceci demande de l’estimation nonparamétrique. En plus la distribution asymptotique χ de ces statistiques peut donner lieu à des approximations peu précises des probabilités dans les queues de la distribution et ceci peut produire des p−valeurs imprécises surtout dans le cas d’échantillons de taille moyenne. Des méthodes alternatives qui ne nécessitent pas de l’estimation de la fonction de “sparsity” sont disponibles, par exemple les méthodes de rang et celles de reéchantillonage pour obtenir des intervalles de confiance qui peuvent être inversés afin de construire des tests. Ces méthodes sont typiquement plus précises que les M−tests standard. Dans cet article on présente une statistique nonparamétrique de pointe de selle qui permet de construire des tests très précis. Elle est asymptotiquement distribuée selon une loi du χ et elle ne nécessite ni de la spécification de la distribution des erreurs ni de l’estimation de la fonction de “sparsity”. La performance de la méthode est démontrée par une étude de simulation, qui montre la robustesse et la précision du nouveau test en comparaison avec les meilleures alternatives disponibles dans la littérature. La revue canadienne de statistique xx: 1–29; 20?? c © 20?? Société statistique du Canada
منابع مشابه
R the Cox Regression Model, Invariance Principles for Some Induced Quantile Processes and Some Repeated Significance Tests @bullet the Cox Regression Model, Invariance Principles for Some Induced Quantile Processes and Some Repeated Significance Tests*
For the Cox regression model, the partial likelihood functions involve linear combinations of induced order statistics. Some in-variance principles pertaining to such linear combinations of induced order statistics are studied and the theory is incorporated in the formulation of suitable repeated significance tests (for the hypothesis of no regression) based on these partial likelihoods. Asympt...
متن کاملUsing quantile regression to create baseline norms for neuropsychological tests
INTRODUCTION The Uniform Data Set (UDS) contains neuropsychological test scores and demographic information for participants at Alzheimer's disease centers across the United States funded by the National Institute on Aging. Mean regression analysis of neuropsychological tests has been proposed to detect cognitive decline, but the approach requires stringent assumptions. METHODS We propose usi...
متن کاملThe Interaction Effects of Good Governance and Public Health Expenditure on Children’s Health Status: Quantile Regression for Upper-Middle Income Countries
Background and Aim: Public health expenditures and the quality of governance are among factors affecting the health status of a population. Therefore, the purpose of this study was to investigate the interaction effects of good governance and public health expenditures on the health status of children in upper-middle income countries. Materials and Methods: This descriptive-analytical applied ...
متن کاملBayesian Quantile Regression with Adaptive Elastic Net Penalty for Longitudinal Data
Longitudinal studies include the important parts of epidemiological surveys, clinical trials and social studies. In longitudinal studies, measurement of the responses is conducted repeatedly through time. Often, the main goal is to characterize the change in responses over time and the factors that influence the change. Recently, to analyze this kind of data, quantile regression has been taken ...
متن کاملTesting Structural Change in Conditional Distributions via Quantile Regressions
We propose tests for structural change in conditional distributions via quantile regressions. To avoid misspecification on the conditioning relationship, we construct the tests based on the residuals from local polynomial quantile regressions. In particular, the tests are based upon the cumulative sums of generalized residuals from quantile regressions and have power against local alternatives ...
متن کامل