Several related documents explore (or will explore) in detail various aspects of Bayes linear methods. In particular, this document accompanies [4], which introduces the basic machinery of the approach. We aim to consider in further documents the implementation of the theory from a computational viewpoint; the elicitation of beliefs; and other aspects of Bayes linear methods. An earlier incarnation of the [B/D] package was considered in [2] and [6]; and a resumé of the approach and the package is given in [7]. More recently, [8] is a technical reference manual, and [5] considers in some detail advanced [B/D] macros used to help analyse an industrial problem.
Here we are concerned with a gentle introduction to some of the features of the approach and to the package used to perform the analyses. The theoretical machinery of the approach is considered in detail in the companion document [4], and this document is intended to provide a tutorial on the interpretation and computation of the material contained therein. For our main illustration we take a genuine problem concerning the diagnoses of diabetes in the elderly. We will assume that we have already undergone the effort necessary for the consideration of all the inputs necessary for our illustration, so that the belief specifications are ready and waiting to be used. This is not to understate the importance of such belief elicitation aspects, but because we concentrate on those features of our approach elsewhere. Similarly, we will not restate the theory and interpretation found in the companion document [4], or various implementational aspects.
This document consists of three parts. In the second part we take a simple (but genuine) problem and show how we apply Bayes linear methods to perform adjustments of belief (informally, the revision process) and to assess the implications of both the belief specifications and the actual observations for the problem. In this way, we explore the central features of the Bayes linear approach.
In part 3 we show how our computer program [B/D] may be used to perform all the analyses considered in part 2, and in so doing we provide a tutored introduction to the use of the [B/D] programming language. In this first part we motivate and introduce the example that we use to illustrate both the methodology and the programming language. It is our intention that the part containing our illustration of the methodology via example stands separate from the final, programming part in case you have no access to (or interest in!) the [B/D] program.