Spatial and temporal variation of mortality and deprivation 2: Statistical modelling

Building on the tabular analyses exemplified in our first paper and widely used in the medical literature, we use generalised linear models to provide a formal, statistical approach to the analysis of mortality and deprivation relationships, and their change over time. Three types of fixed effects model are specified and estimated with the same ward-level data sets for Wales examined in our first paper. They are: Poisson models for analysing mortality and deprivation at a single cross section in time; repeated-measures Poisson models for analysing mortality-deprivation relations, not only at cross sections in time, but also their changes over time; and logit models focusing on temporal changes in mortality-deprivation relationships. Nonlinear effects of deprivation on mortality have been explored by using dummy variables representing deprivation categories to establish the connection between formal statistical models and the tabular approach.