A Semiparametric Estimation Approach for Linear Mixed Models

• Published in: Communications in statistics. Theory and methods Vol. 42; no. 11; pp. 1982 - 1997
• Main Authors: Li, Daniel, Wang, Liqun
• Format: Journal Article
• Language: English
• Published: Philadelphia Taylor & Francis Group 01.06.2013; Taylor & Francis Ltd
• Subjects: Asymptotic properties; Consistency; Economic models; Errors; Estimators; Influence functions; Least squares method; Linear mixed models; Mathematical analysis; Misspecification; Outliers; Redescending M-estimator; Robustness; Statistics; Studies
• ISSN: 0361-0926; 1532-415X
• DOI: 10.1080/03610926.2011.601837

Maximum likelihood approach is the most frequently employed approach for the inference of linear mixed models. However, it relies on the normal distributional assumption of the random effects and the within-subject errors, and it is lack of robustness against outliers. This article proposes a semiparametric estimation approach for linear mixed models. This approach is based on the first two marginal moments of the response variable, and does not require any parametric distributional assumptions of random effects or error terms. The consistency and asymptotically normality of the estimator are derived under fairly general conditions. In addition, we show that the proposed estimator has a bounded influence function and a redescending property so it is robust to outliers. The methodology is illustrated through an application to the famed Framingham cholesterol data. The finite sample behavior and the robustness properties of the proposed estimator are evaluated through extensive simulation studies.