Robust inference in a linear functional model with replications using the t distribution

JOURNAL OF MULTIVARIATE ANALYSIS

卷:160

Article

Galea, Manuel¹  de Castro, Mario² 

[1] Pontificia Univ Catolica Chile, Santiago, Chile

[2] Univ Sao Paulo, Inst Ciencias Matemat & Comp, Sao Carlos, SP, Brazil

关键词: Estimating equations;    Linear functional relationships;    Profile likelihood;    Replicated observations;    Sandwich estimator;   

DOI  :  10.1016/j.jmva.2017.06.008

来源: Elsevier

【 摘 要 】

In this paper, we investigate model assessment, estimation and hypothesis testing in a linear functional relationship for replicated data when the distribution of the measurement errors is a multivariate Student t distribution. For statistical inference, we adopt the unbiased estimating equations approach. The resulting estimator is consistent and asymptotically normal; a closed form expression is also given for its asymptotic covariance matrix. A simple graphical device for model checking is proposed. We also describe how to test some hypotheses of interest on the parameter vector using the Wald statistic. A simulation study is performed to gauge the performance of the estimators and of the Wald statistic. The methodology developed in the paper is illustrated with a real data set. (C) 2017 Elsevier Inc. All rights reserved.

【 授权许可】

Free