Abstract
We compare generalized canonical variate analysis and related metric scaling for simultaneous visualization of multivariate data. We prove that both representations can be quite similar, but under some conditions such as orthogonality, the results can be substantially different. Two real and one artificial examples are given.
Acknowledgments
Work supported in part by grants from Spanish MCT (BFM2000-0801) and Generalitat of Catalonia (2001SGR00067).