Linear Mixed Effect Models With Semiparametric Generalized Skew-Elliptical Random Effects
by
Yanyuan Ma
CRSC, North Carolina State University, Raleigh
Coauthors: Marc G. Genton, Marie Davidian

We relax the standard normal assumption of the random effects in linear random effect models to semiparametric generalized skew-elliptical distributions (SGSE). The SGSE family is able to take into account heavy tail, skewness, and multimodality through the elliptical part, the coefficients in the skewing function part, and a tuning parameter. When the tuning parameter is set to 0, the distribution reduces to an elliptical distribution. SGSE family facilitates simple sampling procedure, which reduces computation complexity for Monte Carlo type algorithms. It also avoids artifactual fluctuations which tend to occur for SNP distributions and mixture distributions.

Date received: August 18, 2003