Linear Mixed Models I

Abstract

Observations often fall into groups or clusters. For example, longitudinal data consist of repeated observations on the same subjects. Hierarchical data sets typically consist of subjects nested in higher level units, such as families or GP practices. In both types of data, we cannot assume that observations on the same subject (or cluster) are independent. Standard methods of analysis such as ANOVA or multiple regression, which assume that observations are independent, are therefore not valid for clustered data. Fortunately, these methods can be extended by explicitely modeling the covariances among observations within a cluster. In this chapter, we discuss how this can be done using linear mixed models.