Linear mixed model with Laplace distribution (LLMM) (Q1706473)

This article proposes assuming multivariate Laplace distributions instead of normal distributions for both the random effects and error terms of linear mixed models. The estimators or the parameters of interest are generated using the EM algorithm, exploring the characterization of the Laplace distribution as a scale mixture of normal distributions. A simulation study is developed, implementing two different scenarios. The first scenario is generated from mixtures of normal distributions and the second one from mixtures of Laplace distributions with and without contamination factors. A real data example of modeling of orthodontic data is presented.