ABSTRACT
The likelihood principle means that likelihood methods can give an efficient way of data analysis if the model is right. Thus, it is important to check the model to verify the analysis of data. This chapter presents Double HGLM (DHGLM) which is a very general class of models for a single response variable. HGLMs have component models for parameters, where only the model for µ allows random effects. In DHGLMs there are many design matrices and parameters to be introduced, which requires some new notation. Modeling of variance has been a special interest in two areas, quality improvement and finance. In the analysis of data from quality improvement experiments, the aim is to minimize variation among products. Rubin and Wu analyzed schizophrenic behavior data from an eye-tracking experiment with a visual target moving back and forth along a horizontal line on a screen.
