CMStatistics 2021: Start Registration
View Submission - CMStatistics
B1771
Title: Diagnostic tools for random effects in general mixed models Authors:  Reza Drikvandi - Durham University (United Kingdom) [presenting]
Abstract: Mixed models are frequently used for the analysis of longitudinal, multilevel, clustered and other correlated data. They incorporate subject-specific random effects into the model to account for the unknown between-subject variability as well as the within-subject correlation. Since random effects are latent and unobservable variables, it is difficult to assess the random effects and their assumed distribution. There are two main challenges when working with random effects. The first challenge is to decide which random effects to include in the model. The second challenge is to check the appropriateness of the assumed distribution for random effects, which is a more difficult task. We first introduce permutation and Bayesian tests for the inclusion or exclusion of random effects from the model. We then present a likelihood-based diagnostic tool to check the adequacy of random-effects distribution. The proposed diagnostic tools can be used to assess random effects in a wide class of mixed models, including linear, generalised linear and non-linear mixed models, with univariate as well as multivariate random effects. The methods are illustrated via real data applications.