ABSTRACT

This chapter explores two types of questions how can EBVTM be simplified further without additional loss of information and when is a pair of experiments a jointly admissible reduction of EAVS. The first question concerns the admissible reduction of EBVTM; having accepted a loss of information in the reduction of EAVS to EBVTM, one seeks to further simplify EBVTM without additional loss of information. The second kind of question is rather different in nature. The second question decomposed the learning process either on the sample space or on the parameter space. The chapter shows that because a tnarginal experiment has exactly the same structure as an unreduced experiment — viz., a unique probability II on a product space — the concepts of ancillarity and of sufficiency are essentially the same.