General statistical methodology

A particular aspect of general methodological research has been the study of key inferential ideas behind attempts to draw causal conclusions from observational or experimental longitudinal data. In a typical situation, several individuals belonging to a study population are followed over time, with a sequence of covariate measurements along with values of 'treatment' variables that will be interpreted as causes, being registered for each individual. Particular attention in this context is given to the potentially important problem of confounding, and to conditions under which the size of causal effects can be estimated by statistical methods. Our approach to this inferential problem is Bayesian, and uses predictive distributions as summary measures of the causal effect.