percentage of cross-block covariance computation
giorgio.arcara
Posted on 11/30/11 08:22:46
Number of posts: 6
Hello,
I'm trying to compute the cross-block covariance by my self to better understand the math beyond PLS.
I run a mean-centering PLS on my data. According Caplan et al. (2008) the percentage of cross-block covariance is "the singular value divided by the sum over all singular values".
However if I do this computation by myself the results are quite different.
After loading the PLS results by matlab prompt
I tried to compute cross-block covariance as
s/sum(s)
my results however are quite different from the percentage of cross-block covariance reported in Design scores plot.
Am I missing something?
Thanks!!
Giorgio