ABSTRACT
In the previous chapters we have looked at a range of methods of building models of distributions. In this chapter we look at the identification of suitable models for a given set of data. There are a number of aspects of identification. There is an initial need to understand the context in which the data arose, then the data needs to be pictured in as many ways as possible. This is done mainly by using graphical techniques, many of which we have already discussed. We may need some numerical information about the data. On the basis of these studies we may have ruled out some possibilities and highlighted others. The next stage is to do a range of comparisons with a list of candidate models. This may involve separate studies of the tails of the data. From this stage we will hopefully home onto a few real candidates. These will need to be compared with the data in detail. It is advisable at this last identification stage to select more than one model. Having obtained “the chosen few,” these are fitted to the data, using the methods in Chapter 9. The methods of validation, discussed in Chapter 10, are then used to finally decide on the model.
