Building graphical models

The design of a graphical model usually takes into consideration two different aspects. First of all, we need to identify the variables involved in the model. The variables can refer to facts we can observe or measure, such as a temperature, a price, a distance, a number of items, an interval of time, or any other value. A variable can also represent a simple fact that can be true or false.

At the same time, and this is why we're building such models, the variables can capture parts of the problem that we cannot directly measure or estimate but that are related to the problem. For example, a physician is able to see and measure a set of symptoms for a patient. However diseases are not facts that can be directly observed. They can only be deduced from the observation of several symptoms. Let's take the common cold. It is natural to say that someone has a cold, and in practice everybody understands what it means. However, there is no such thing such as a cold, but...