ABSTRACT
This chapter is concerned with the application of hierarchical priors to regression models for univariate metric and discrete responses, where the observation units are non-nested but may be spatially or temporally configured. Nested data applications are considered in Chapters 6 and 8. Particular applications involving latent responses or random effects are when such effects are used:
1. to improve model fit in general linear models in line with distributional assumptions;
2. to generate latent responses on a different scale to the observations;
3. to demonstrate heterogeneity in regression relationships or variance parameters over exchangeable sample units;
4. to represent random regression effects and correlated regression errors for responses structured in time or space.
