Date:         Fri, 18 Apr 1997 21:28:55 GMT
Reply-To:     David Nichols <nichols@SPSS.COM>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         David Nichols <nichols@SPSS.COM>
Organization: SPSS, Inc.
Subject:      Re: mean centering

In article <33518B95.5396@overnet.com.ar>,
Hector E. Maletta <hmaletta@overnet.com.ar> wrote:
>bryan scott lilly wrote:
>>
>> Can any list members advise me?  I have several data sets.  For each data
>> set, I wish to mean center all of the variables, replacing the original,
>> raw values, with mean centered values.  I have access to SPSS 7 and 7.5.
>> Is there a simple technique for accomplishing this task?
>>
>> I realize that I can perform this task by using multiple transform
>> statements, but I'm looking for a simpler solution.
>>
>> Also, does SPSS have multivariate outlier detection capabilities that do
>> not require any dependent variable declaration?  My reason for wanting to
>> mean center my matricies (above) is to follow Bollen's (Structural
>> Equation Models, 1991) prescription for multivariate outlier detection.
>>
>> Bryan Lilly
>> Indiana University
>
>
>The DESCRIPTIVES procedure (to be find under /STATISTICS/SUMMARIZE in
>the menu system) creates Z variables (mean-centered and measured in
>standard-deviation units), which is not exactly what you want but might
>be useful. You can create all the variables at once with this procedure,
>and delete older variables afterwards (e.g. when you save the file).
>
>
>Hector Maletta
>Universidad del Salvador
>Buenos Aires, Argentina
 
If by multivariate outlier detection capabilities you mean distances in
multivariate space, you can compute either leverages of Mahalanobis
distances in the REGRESSION procedure (these are constant multiples of
one another). While a dependent variable is required in REGRESSION, it
has no effect on these values, which are solely functions of the right
hand side of the regression equation.
 
--
-----------------------------------------------------------------------------
David Nichols             Senior Support Statistician              SPSS, Inc.
Phone: (312) 329-3684     Internet:  nichols@spss.com     Fax: (312) 329-3668
-----------------------------------------------------------------------------