Our second example extends the use of portraying partial adjustments on
influence diagrams, and introduces the notion of path correlation.
Consider the model (1), and suppose that we observe runs
of the experiment in batches with three observations of both product
yields in each batch. Thus batch 1 consists of the observations
, and so forth. We will adjust the
unknowns by each of the four batches sequentially. (Where
experimentation is very expensive, it is obviously useful to identify
the point at which further experimentation is expected to contribute
little extra knowledge.)