Date:         Wed, 30 Jul 2008 16:57:01 +0200
Reply-To:     Marta García-Granero <mgarciagranero@gmail.com>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Marta García-Granero <mgarciagranero@gmail.com>
Subject:      Re: Parallel Analysis
In-Reply-To:  <9002B8778A49FE46918099941662D2EE1C0884@CTSINGGNSXUA.cts.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed

Manmit.Shrimali@COGNIZANT.COM escribió:
> Hi ALl:
>
> I have recently heard lot on parallel analysis. I was wondering if anyone can share their experience whether running PA was any help determing number of factors? If spss generates screen plot and eighenvalue so wondering what is the value of PA.
>
The number of factors to be retained can be determined by several
procedures, and there is no clear consensus to which one is the best. It
depends on sample size, number of variables...

- Cattell's criterion uses the scree plot (not "screen"). Sometimes it
retains too few factors
- Kaiser's criterion, the oldest, selects those factors with eigenvalues
over 1. It has been much criticized and it's considered to yield
non-parsimonious solutions, with too many factors retained
- Velicer's MAP test and Horn's parallel analysis give in general good
results
- Percent variance explained criterion
- Others..

There are a lot of other details to be taken into account, like if
different solutions with different number of factors are interpretable
or not, a priori hypotheses...

Take a look at this page (accurate googling is a powerful tool for
finding answers): http://www2.chass.ncsu.edu/garson/pa765/factor.htm#facnum

HTH,
Marta García-Granero

--
For miscellaneous statistical stuff, visit:
http://gjyp.nl/marta/

=====================
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
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD