|
Abstract:
|
The linear mixed effects model is widely used in the analysis of repeated measures data with continuous response. Two of the key assumptions for the linear mixed model are that the random effects and the errors are normally distributed. In practice, these assumptions are not easy to check. What are the impacts if these assumptions are not satisfied? Several papers have studied the robustness of the inference on the fixed effects with misspecified random effects distribution (Butler and Louis, 1992, Verbeke and Lesaffre, 1997) or error distribution (Jacqmin-Gadda et al.,2007). While the existing work focused on fixed effects estimate, control of type I error and asymptotic property, we further studied the impact of the misspecified error distribution by comparing power as well as type I error for the inference of fixed effects by simulations for data with small/medium sample sizes.
|
