ABSTRACT
This chapter explains how random effects models can be used within Generalized additive models for location scale and shape (GAMLSS). It introduces the different ways in which random effects can be fitted within GAMLSS models. The chapter examines the advantages and disadvantages of such modelling. It provides examples demonstrating the different approaches. Random effect methodology is a powerful tool in statistics which was originally introduced to deal with group data. Random effects models deal with correlation between individual observations by introducing different sources of variation in the data. The functions random() and re() are based on penalized quasi likelihood methodology, and can be used like all other additive smoothing terms within GAMLSS to model any or all of the parameters of the response distribution, using normal random effects. The nonparametric random effect model can be applied to one or more of the distribution parameters.
