GRAPHALS

There is a lot of material in applied math (computer science, computational combinatorics) or drawing graphs elegantly in the plane (or n-space). As we know, the indicator matrix defines a bipartite graph on the objects and categories (which some additional structure because of the variables). One way to think of MCA is as a "string" algorithm to draw a graph -- we want the lines to be as short as possible. Other Gifi techniques can be fitted into this picture: they impose constraints on the location of the category points. By going through the graph drawing literature, we may find interesting alternatives (in particular to the disgusting X'X=I constraints).

UCLA Department of Statistics
Last updated: 12-Oct-2000
Access count is: 2667, since 01-Jul-2001
Maintained by: Jan de Leeuw [deleeuw@stat.ucla.edu]