ABSTRACT
In the previous three chapters we have reviewed likelihood inferences about fixed parameters. There have been several attempts to extend Fisher likelihood beyond its use in parametric inference to inference from more general models that include unobserved random variables. Special cases are inferences for random-effect models, prediction of unobserved future observations, missing data problems, etc. The aim of this chapter is to illustrate that such an extension is possible and useful for statistical modelling and inference.
