ABSTRACT
This chapter focuses on delineating the mixture of experts modelling framework and demonstrates the utility and flexibility of mixture of expert’s models as an analytic tool. It introduces the generic mixture of experts (ME) framework and describes approaches to inference for ME models. The chapter discusses a broad range of illustrative data analyses and provide an overview of existing software packages which fit ME models. It also discusses identifiability issues for ME models and deals with some the benefits and issues of the ME framework, and of some areas ripe for future development. A simple simulated data set is employed to introduce the ME framework. The exact manner in which an ME model is estimated again depends on the nature of the ME model and the outcome variable. The expectation-maximization algorithm provides an efficient approach to deriving maximum likelihood estimate in ME models.
