Practise

We begin with a look at one of the most famous datasets in all of Statistics, Fisher's iris data. The problem is to devise a "rule" based on the measurements to decide what type of flower we have.
In iris.fun() we use all of the ideas previously considered (with the exception of quadratic regression). We see that the error rate is on the order of a few percent, most methods miss-classify 3 of the flowers.

Another interesting dataset is for the Ratings of Painters. Again we use all the methods to discriminate between the Schools, in painters.fun(). Here the mcr is very high, no better than 40% or so for even the best methods. This is not surprising because we have very little data and 8 groups, and the ratings themselves are subjective. QDA actually does not work at all, with an error message pointing out that there is not sufficient data.