Nonlinear Canonical Correlation

[dincooo]

Hi everyone,

Do you know any SAS procedure for "nonlinear canonical correlation"?

If yes, could you please help me to reach the related documents or examples?

There is a procedure related to that (PROC CANCORR) but I could not find anything about nonlinearity.

The purpose of canonical correlation analysis (Mardia, Kent, and Bibby; 1979) is to explain or summarize the relationship between two sets of variables by finding a small number of linear combinations from each set of variables that have the highest possible between-set correlations. Plots of the canonical variables can be useful in examining multivariate dependencies. If one of the two sets of variables consists of dummy variables generated from a classification variable, the canonical correlation is equivalent to canonical discriminant analysis (see Chapter 27, The CANDISC Procedure ). If both sets of variables are dummy variables, canonical correlation is equivalent to simple correspondence analysis.

Many thanks & Best Regards

Onur

[SteveDenham]

I really don't have an answer on this, but let's see what can be done.

First, what nonlinear functions are you considering as basis functions for your two groups of data?  There might be something available in the ETS package that could help (PROC MODEL, PROC COPULA or PROC SPECTRA).  Second, would something like a wavelet decomposition allow you to identify the degree of relationship?

Steve Denham