Robust inference in linear mixed model with skew normal-symmetric error

Abstract

Linear mixed effects models with general skew normal-symmetric (SNS) error are considered and several properties of the SNS distributions are obtained. Under the SNS settings, ANOVA-type estimates of variance components in the model are unbiased, the ANOVA-type F-tests are exact F-tests in SNS setting, and the exact confidence intervals for fixed effects are constructed. Also the power of ANOVA-type F-tests for components are free of the skewing function if the random effects normally distributed. For illustration of the main results, simulation studies on the robustness of the models are given by comparisons of multivariate skew-normal, multivariate skew normal-Laplace, multivariate skew normal-uniform, multivariate skew normal-symmetric, and multivariate normal distributed errors. A real example is provided for the illustration of the proposed method.

Keywords

Skew normal-symmetric (SNS) / ANOVA-type F-test / mixed effect

References

[1] Arellano-ValleR B, GentonM G. Fundamental skew distributions. J Multivariate Anal, 2005, 96: 93–116

[2] AzzaliniA. The skew-normal distribution and related multivariate families (with discussion). Scand J Stat, 2005, 32: 159–188

[3] AzzaliniA, Dalla ValleA. The multivariate skew-normal distribution. Biometrika, 1996, 83: 715–726

[4] GentonM G. Discussion of “The Skew-normal”. Scand J Stat, 2005, 32: 189–198

[5] GuptaA K, ChangF C. Multivariate skew-symmetric distributions. Statist Probab Lett, 2003, 16: 643–646

[6] HuangW J, ChenY H. Quadratic forms of multivariate skew normal-symmetric distributions. Statist Probab Lett, 2006, 76: 871–879

[7] KhuriA I, MathewT, SinhaB K. Statistical Tests for Mixed Linear Models. New York: John Wiley & Sons, 1998

[8] LachosV H, DeyD K, CanchoV G. Robust linear mixed models with skew-normal independent distributions from a Bayesian perspective. J Statist Plann Inference, 2009, 139: 4098–4110

[9] LachosV H, GhoshP, Arellano-ValleR B. Likelihood based inference for skew-normal independent linear mixed models. Statist Sinica, 2010, 20: 303–322

[10] LairdN M, WareJ H. Random effects models for longitudinal data. Biometrics, 1982, 38: 963–974

[11] MaY, GentonM G. Flexible class of skew-symmetric distributions. Scand J Stat, 2004, 31: 459–468

[12] NadarajahS, KotzS. Skewed distributions generated by the normal kernel.Statist Probab Lett, 2003, 65: 269–277

[13] SearleS R, CasellaG, McCullochC E. Variance Components. New York: Wiley, 1992

[14] WangS G, ChowS C. Advanced Linear Models: Theory and Applications. New York: Marcel Dekker, Inc, 1994

[15] WuM X. Introduction to Linear Mixed Models. Beijing: Science Press, 2013 (in Chinese)

[16] WuM X, WangS G. Simultaneous optimal estimates of fixed effects and variance components in the mixed model. Sci China, Ser A, 2004, 47: 787–799

[17] WuM X, ZhaoJ, WangT H, ZhaoY. The ANOVA-type inference in linear mixed model with skew-normal error. J Syst Sci Complex, 2017, 30: 710–720

[18] YeR D, WangT H. Inferences in linear mixed model with skew-normal random effects. Acta Math Sin (Engl Ser), 2015, 31: 576–594

[19] YeR D, WangT H, GuptaA K. Tests in variance components models under skewnormal settings. Metrika, 2015, 78: 885–904

RIGHTS & PERMISSIONS

2017 Higher Education Press and Springer-Verlag GmbH Germany