Re: factor analysis

Fri, 8 Dec 1995 14:48:24 +1100 (EST)

>I have a question about reporting statistical information in a manuscript. I
>performed a principal components factor analysis on the ways subjects rated
>a vignette character along 40 bipolar dimensions. I used an ablique rotation.

Did you use oblique because you thought that the factors should be correlated?
If so thats ok. If you thought that they should be independent then you
should use

>What I am uncertain of (nobody seems to be able to answer this w/certainty)
>is what data I need to report in a table presenting the factor analysis'
>results. Do I simply list the factor loadings, or do I also need to include
>communality scores (I know I would need to include communality scores if it
>were an orthogonal rotation, but do I still need to do so when using an
>oblique rotation?)?

Forget communality scores report eigenvalues % of variance by factor and
factor loadings above .4 or .5

Be careful when using any rotation method. You will need to recalulate the
eigenvalues and thus the percentage of variance accounted for by each
factor. Do this by squaring the loadings on each factor and sum them up.
You can then realloacte the % of variance for each factor on a prorata basis.

Look at previous issues of the journal but as I indicated I do not think
that communalities are necessary. Read the SPSS user guides for info on the
significance of communalities.

Cheers and happy Christmas from downunder

Alastair Anderson
>Any assistance would be much appreciated! I am trying to get the manuscript
>containing this data sent off for review before I go on Winter Break!
>Thanks everyone!
>Jonathan D. Raskin, Ph.D.
>Dept. of Psychology
>Tennessee State University
>3500 John A. Merritt Blvd.
>Nashville, TN 37209-1561
>(615) 963-5158
>e-mail: raskinj@HARPO.TNSTATE.EDU