Date: Sun, 23 Jan 2011 21:46:45 +0100
Reply-To: Tanja Gabriele Baudson <firstname.lastname@example.org>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Tanja Gabriele Baudson <email@example.com>
Subject: Reliability analyses of IQ tests
Content-Type: text/plain; charset=ISO-8859-1; format=flowed; delsp=yes
when running a reliability analysis, I get the well-known "The
determinant of the covariance matrix is zero or close to zero"
message. I understand what this message means (thanks to the SPSSX
archives) but am wondering about two points, on which your input would
be greatly appreciated:
* The result is that some values cannot be computed (such as squared
multiple correlations). However, for the other indices which are not
based on the inverse matrix and are therefore computed and given in
the output (Cronbach's Alpha, Corrected Item-Total Correlations etc.),
this should not be a problem, thus I might use them as they are. Or am
I mistaken? My sample size is N > 500.
* In addition, I have a conceptual question. My test is an IQ test,
and therefore the items (tapping the g factor in some way or another)
SHOULD actually correlate highly, shouldn't they?
In my example, I have eight subtests assessing different aspects of
figural, verbal, and numerical reasoning. For me, it is quite obvious
that items from one and the same subtest, thus tapping the same
ability (e.g., solving matrices), correlate highly when I compute the
internal consistency across the entire scale of 36 items ...
Thanks very much in advance
Tanja Gabriele Baudson
FB I Psychologie
Hochbegabtenforschung und -förderung
To manage your subscription to SPSSX-L, send a message to
LISTSERV@LISTSERV.UGA.EDU (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
For a list of commands to manage subscriptions, send the command