Date: Wed, 14 Sep 2005 12:06:09 -0300
Reply-To: Hector Maletta <hmaletta@fibertel.com.ar>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Hector Maletta <hmaletta@fibertel.com.ar>
Subject: Re: averaged scales vs factor scores
In-Reply-To: <20050914143357.152.qmail@web34503.mail.mud.yahoo.com>
Content-Type: text/plain; charset="us-ascii"
Plance,
Factor scores are measuring a non observable trait, whose unit of
measurement is unknown. Therefore it is assumed to have zero mean and unit
standard deviation, and all individual scores respond to that convention.
The [population] mean and std dev of factor scores are taken to coincide
with the mean and std dev of the sample on which factor scores are
estimated. Therefore, if you put the same subject in two different samples
with the same test results in both cases, that subject would get different
factor scores, because the scores would be expressed in each case in
relation to the mean and standard deviation of the sample where the subject
is being included. The score is like a percentile grading, since it depends
on the scores of other people in the sample. A subject may be in the top 10%
in one sample and the bottom 10% in another.
However, you are not usually interested in each subject individually, but in
the distribution of the scores and their correlation with other variables.
That is not supposed to vary from sample to sample, besides normal sampling
fluctuations.
Analysis relating factor scores to other variables should not be affected.
If you find that factor scores are, for instance, related to income or a
socioeconomic status scale, that finding would not depend on the unit of
measurement of the variables involved. So if subject A (who has high income)
passes from top to bottom factor score after moving to another sample, some
subject B (also with a high income) would be doing the opposite, and the
distribution would probably not change (if the two samples are true random
samples drawn from the same population). Besides, it would be rather
unlikely that two random samples differ by so much that the same subject is
in the top of one and the bottom of the other.
Hope this helps.
Hector
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU]
> On Behalf Of Plance Debouver
> Sent: Wednesday, September 14, 2005 11:34 AM
> To: SPSSX-L@LISTSERV.UGA.EDU
> Subject: averaged scales vs factor scores
>
> I posted this with no lucky. How do I turn factor scores
> based on rotated loadings of FA procedures (scores obtained
> using regression method)into useful values for further
> analysis. I would like to compare 5 groups of respondents.e.g
> nurses, doctors, Lab
> technicians, students. I am told that factor scores
> are not comparable across studies (they are sample
> specific) but summated or averaged scores are.
>
> Any recommendation for types of further analyses on factor
> scores or summed or averaged scales would be appreciated.
>
> plance.
>
>
>
> __________________________________
> Yahoo! Mail - PC Magazine Editors' Choice 2005 http://mail.yahoo.com
>
> __________ Informacisn de NOD32 1.1217 (20050914) __________
>
> Este mensaje ha sido analizado con NOD32 Antivirus System
> http://www.nod32.com
>
>
| 
