This tutorial document illustrates two important aspects of Bayes linear methodology: the generation of Bayes linear influence diagrams, and the exploitation of exchangeability. In each case we give examples and show how they are analysed using [B/D]. Section 2 considers the construction of influence diagrams for a problem in which two correlated linear regressions are used to model a certain chemical process. Section 3 shows how we model exchangeable structures, and how [B/D] can be used to exploit the exchangeability to suggest sampling strategies under various circumstances.
Two related technical reports are [6] and [15]. The former introduces the basic machinery of Bayes linear methodology and in particular describes the objectives of such analyses and the notation used. The latter shows how Bayes linear methods are applied to a simple example, and then how the [B/D] language can be used to carry out the analysis. This technical report assumes familiarity with its two predecessors. Detailed instructions on constructing influence diagrams using [B/D] can be found in [16].