ABSTRACT
This chapter provides some basic properties of sequences of conditional independence in the context of a general probability space. It describes the formal framework of a sequential Bayesian experiment and the different levels of analysis that appear to be relevant. These different levels of analysis lead, in general, to non-equivalent conditions for admissible reduction. The relationships between different forms of non-causality have been presented in Florens and Mouchart while the role of non-causality in Markovian models has been analyzed in Florens, Mouchart and Rolin. The new problems concern the relationships between admissible reductions at different levels of analysis. It investigates the relationships between initial and sequential admissibility of reductions.
