How to get a redundancy index when performing canonical correlation analysis in SPSS? I am trying to perform a canonical correlation analysis to investigate the relationship between attitudes (14 variables), perceived consumer effectiveness (6 variables) AND Intention to dine (DV) 14 variables. However when SPSS generates the Manova Output there are no  tables on redundancy index and as far as I understand  this index is important to report as it explains how good are the IVs as the predictors of the variance in DV. 
This is the script that  i am using for the test:
manova Attitudes PCE with Intention
/ discrim all alpha(1) 
/ print=sig(eigen dim).

Thank you!
Oksana
 A: It's been a while since I've run canonical correlation in SPSS.
Perhaps have a look at the cancorr macro. It's an alternative way of running a canonical correlation, and from memory it provides slightly different output.
Running Cancor
Here's some info on running cancorr, extracting the important bit:
"Note that the location script required to run CANCORR changes between versions and installations of SPSS. David Garson sets out the following code template:"
INCLUDE 'c:\Program Files\SPSS\Canonical correlation.sps'.
CANCORR SET1=varlist/  SET2=varlist/.

Additional Reference


*

*Here's some more information from a book chapter by V.K. Bhatia
A: As it is not really difficult to import SAV dataset in R nowadays, with e.g.,
library(foreign)
df <- read.spss("yourfilename", to.data.frame=TRUE)

you can check your SPSS results against one of the R packages that allow to perform CCA (see the CRAN Task View on Multivariate or Psychometrics analysis). In particular, the vegan package offers an handy way to apply CCA and has nice graphical and numerical summary through the CCorA() function.
Also, note that redundancy indexes apply onto one block of variables, conditional on the other block (hence the distinction you'll find in the aforementioned function between Y|X and X|Y); they are intended to provide a measure of the variance of one set of variables predicted from the linear combination of the other set of variables. However, in essence CCA consider that you have two sets of measures that play a symmetrical role. They are both descriptions of the same individuals or statistical units. If your blocks really play an assymmetric role--that is you have a block of predictors and a block of response variable--then you're better using PLS regression.
