Date:         Fri, 21 Apr 2000 13:50:41 -0400
Reply-To:     Callaghan <igmusic@EROLS.COM>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Callaghan <igmusic@EROLS.COM>
Subject:      canonical correlation
Content-Type: text/plain; charset=us-ascii

I ran a canonical correlation to answer my dissertation research
question using the SPSS 10.0 program. I have a couple of questions
regarding the print-out that was generated by this statistic. First of
all, according to the readings I have been doing on canonical
correlation, it appears that the standardized canonical coefficients for
sets 1 and 2 are the "beta weights" for the variates and when squared,
these canonical coefficients can be interpreted as variance accounted
for. The total variance accounted for by a particular variate is
obtained by adding the variances for each variable within a variate and
then dividing by that number of variables. Therefore, the variance
accounted for by a particular variate is not obtained by squaring the
canonical correlation for that variate. Is my interpretation correct? My
second question is in relation to the canonical loadings portion of the
print-out. How are these loadings interpreted? Or are they just the
correlations that are used to determine redundancy measures? My final
question is in relation to the redundancy analysis. It appears that the
redundancy measure that should be reported in my data analysis chapter
is the proportion of variance of one set explained by the other set. I
read that the coefficients generated by this analysis need to be
multiplied by the canonical correlation coefficient for that particular
variate in order to obtain the redundancy measure. Is my interpretation
correct? I hope someone out there can help me because I haven't been
able to find some one that can.
Donna Callaghan