Building the Model

Summary

This chapter closes this part dedicated to the construction of a specific model.

It first tries to motivate the reader of the interest of building such a model.

Of course model building assumes that data are available, that they have been analyzed in order to discover any potential problem, that correlations have been studied, etc.

It also assumes that the difference between “sample” and “population” has been clearly understood.

Once everything has been studied, model building is a rather easy task. The role of the model builder is then to make several decisions.

All decisions are based on what is expected for the formula: When it is going to be used? What is the precision which is looked for? Is the information available in terms functional or physical?, etc.

The first decision refers to the variables that should be included in the model. Some comments are made in this chapter, in addition to everything which has been already said.

The second decision is the kind of formula which has to be selected. Among the infinite number of types of formulae, the best compromise has to be found between a complex formula which completely describes the data and the simple one which is easy to handle.

The third decision is the choice of the metric to be used. This choice depends on several things from an a priori knowledge of the scattering of the data.

When the formula has been built, its quality should be checked and quantified, one of the reasons for this quantification being to be able to compare different solutions. The second one, maybe more important, is to communicate the information to potential users.