Posted by
statisticsdoc on
Sep 18, 2006; 2:08am
URL: http://spssx-discussion.165.s1.nabble.com/Principal-component-analysis-with-clusters-tp1070978p1070979.html
Stephen Brand
www.statisticsdoc.com
Dear Mr. Meyer,
Are you interested in differentiating the groups according to the factor
scores? In that case, you may want to use the factor structure from all the
cases to compute factor scores, and consider: a.) a simple plot of the
factor scores for each group; and b.) a discriminant function analysis that
constructs linear combinations of the factor scores that differentiate the
factors.
If you are interested in whether there are actually different factor
loadings in the different groups, then that is a different question. The
issue of whether factor loadings differ between groups is an issue of
factorial invariance. You would not compute and overall factor solution for
all cases. Instead, you would want to use AMOS or your favorite structutal
equation modeling software to test a confirmatory model in which the factor
loadings are constrained to be equal across groups versus one in which the
loadings are allowed to differ across groups. Examine whether the model has
poorer fit when the loadings are constrained to be equal. If the
unconstrained model has better fit, consider whether the loadings are
substantively different if they are allowed to vary.
HTH,
Stephen Brand
For personalized and professional consultation in statistics and research
design, visit
www.statisticsdoc.com
-----Original Message-----
From: SPSSX(r) Discussion [mailto:
[hidden email]]On Behalf Of
Luca Meyer
Sent: Sunday, September 17, 2006 5:12 AM
To:
[hidden email]
Subject: Principal component analysis with clusters
We have a dataset with some 20 numerical variables and would like to derive
PC for it. The dataset is composed of 1000 cases divided into 5 groups of
200 cases each. We would like to study the different loadings for each group
on the common factors. In other words, we were thinking to (1) derive a
common factor structure using the 1000 cases, (2) somehow study the
difference between groups according to such loadings.
How would you proceed? We were thinking of FACTOR but then are not sure how
to relate groups to the common factors...
Thank you,
Mr. Luca MEYER
www.lucameyer.com
PS: We licence only SPSS Base, do we need more advanced packages?