A traditionale supervised method will be introduced as this last method for the nominal domain. First of all the successful use of Shannon's entropy (though not often as such recognized) is demonstrated and the potential use of supervised methods for unsupervised structure-finding is shown.
Other supervised, nominal methods can be derived from the supervised, continuous domain. One example is the ``Logit''-model which can be described as Logistic regression for nominal variables. Instead of using a linear combination of the variables it just uses effect variables for each nominal value (similar to ANOVA and log-linear models).