ABSTRACT
Within what limits is it reasonable to suppose the parameter (3 to lie? Are the data consistent with a particular theoretical value for, say, the parameter 13? Let Y r be a further observation assumed to be obtained from the same model and parameter values as the original data, but with
E (Y f ) = 7 + /Jx+. Within what limits is Y 'l expected to lie? In the theoretical discussion, it will be usual to take a family of
models as given. The task is to formulate meaningful and precise questions about such models and then to develop methods of analy sis that will use the data to answer these questions in a sensitive way.