ABSTRACT
D. G. Clayton et al. have stressed the usefulness of including spatial correlation terms in models to make allowance for confounder and ecological biases and so there are reasonably strong arguments for always including contextual terms that have spatial structure. There are many other application areas where spatial context can be useful. One such area that is closely related to longitudinal analysis is the analysis of repeated events. The method is an extension of longitudinal and survival analysis where instead of making one measurement at different times or observing a single time-to-endpoint, the time period is fixed and within that period a sequence of events concerning an individual are observed. As most survival data is observed at the individual unit level there could be individual covariates, random effects or contextual effects relating to the individual. Hence for a parametric survival model, at the first level of the hierarchy, the data model consists of an endpoint distribution.
