Date:         Thu, 31 Mar 2005 21:24:28 -0500
Reply-To:     Art@DrKendall.org
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Art Kendall <Art@DRKENDALL.ORG>
Organization: Social Research Consultants
Subject:      Re: PCA versus FA
Comments: To: Jeff Stuewig <jstuewig@gmu.edu>
In-Reply-To:  <5.1.1.6.0.20050331195827.00b146d8@mail.gmu.edu>
Content-type: text/plain; charset=ISO-8859-1; format=flowed

One way in which these two forms of factor analysis Principal Components
and Principal Axis differ is how much of the total variance of the items
they try to account for.  Items are considered to be compose of 3 parts,
common variance, unique variance (i.e., specific to the items), and
random error variance. In PC you try to account for both the common and
unique variance. (Leave ones in the diagonal.) In PAF you are only
interested in the common construct underlying a set of items. (So you
put some estimate of  communality on the diagonal, ususally the squared
multiple correlation with the set of other items.)

I suggest you google to find out how structures are viualized in
structural equations modelling.

Art
Art@DrKendall.org
Social Research Consultants
University Park, MD  USA
(301) 864-5570

Jeff Stuewig wrote:

>        I've been having difficulty explaining Principal Component
> Analyses to
> some students.  If I understand it correctly (and that is a big if), PCA
> assumes no error in measurement.  So would it be correct if I was drawing
> this relationship that the arrows from the observed variables will point
> toward the different components?  While in Factor Analyses (principal
> axis)
> the arrows go from the latent constructs (factors) to the observed
> variables?  Thanks for any help.
>
>                Jeff
>
>