A projection pursuit method on the multidimensional squared contingency table (Q1887235): Difference between revisions

For multidimensional contingency tables with equal number of categories in each dimension the technique of projection pursuit is developed based on the affine geometry properties of these tables and the Radon transform. It allows to find the ``interesting projections'', i.e., the marginals with nonuniform structure. A graph visualization is proposed for this method which is similar to the discrete version of Andrews' curve. Results of analysis of market basket data and data on the style of Plato's ten works are presented.