Fisher's Discriminant Functions

There are many ways in which the classification of the samples can be accomplished. Fisher's discriminant functions is the simplest method.

To assign an object to one of the pre-defined groups, we use the classification equations generated by DA to give a score for a sample. These equations have the form:

where cj is a classification function coefficient,  j is the group and p the number of variables.

Fisher's discriminant functions give the coefficients, c, in the above equation.

Using the example above, from the Romano British pottery demo data set, the functions for each group are:

Caldicot:

Llanederyn

New Forest

A sample is allocated to the group for which it has the highest classification score.

For example, a sample with 14% Al, 7% Fe, 4% Mg, 0.1% Ca and 0.5% Na gives

so

and in similar fashion

This sample is therefore allocated to the Llanederyn group, as it has the highest score.