Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: RE: Factor Analysis and correlation table


From   "Nick Cox" <n.j.cox@durham.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   RE: st: RE: Factor Analysis and correlation table
Date   Mon, 28 Apr 2008 17:49:26 +0100

Please say if you are using factor analysis or principal component
analysis. 
For many people, if not all, these are quite distinct techniques,
despite the relations between them. 

Even better, please say exactly what Stata commands you are using. 

Compare advice for all: 

"Say exactly what you typed and exactly what Stata typed (or did) in
response. N.B. exactly!"

<http://www.stata.com/support/faqs/res/statalist.html#stata>

Thanks, Nick
n.j.cox@durham.ac.uk 

Timura, Catherine A PhD 

Sorr, it was my first post and I needed to clarify things.
I did the principal components analysis over my agreement table in both
STATA and Systat.  While the eigenvalues are identical, the factor
loadings are different.  It seems that Stata is doing an additional
correlation over the correlation table, significantly dropping the
factor
loadings.  My question is whether I can change the way that Stata
generates the factor loadings to prevent this from happening, or if
other
people have noticed this problem.

> The principal diagonal of a correlation matrix in my experience
contains
> 1s, as it includes the correlations between each variable and itself.
>
> For principal component analysis, that would seem the obvious choice
> too.
>
> If there is a reason for something else, perhaps as a consequence of
> using some flavour of factor analysis (as your title, but not your
> content, implies) then I think you will have to specify what you want
in
> the principal diagonal.
>
> In short, either the diagonal contains 1s, or there is a reason for
your
> diagonal to contain something else, and you will have to specify what
> that is.
>
> Timura, Catherine A PhD
>
> I am using stata 10 for mac. I want to do a principal component
analysis
> over a correlation table (and not have the program do an additional
> correlation over the data).  While the option exists in the menu, it
> does
> not allow the diagonal to be blank.  Please advise on how to allow for
> the
> diagonal to remain blank, to avoid having it part of the analysis.
> We are doing cultural consensus analysis over hand-drawn agreement
> tables,
> as correlations underestimate agreement between individuals.

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index