Date:         Tue, 23 Aug 2005 15:50:43 -0400
Reply-To:     qq.consulting@sympatico.ca
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Susan Elgie <qq.consulting@sympatico.ca>
Subject:      Re: Residual correlation graphs
Comments: To: Spousta Jan <JSpousta@CSAS.CZ>
Content-Type: text/plain; charset=iso-8859-1

Hi Jan,

This is EXACTLY what I need.  Thank you immensely.  I made just one tiny adjustment, to remove redundant correlations.  The solution is elegant and just made my day.  Thanks too to Gene Maguin for encouragement.  This is a wonderful list!

All best wishes,

Susan


> Hi Susan,
>
> Is this what you need?
>
> * Get an example data.
> GET  FILE='C:\Program Files\SPSS\GSS93 subset.sav'.
>
> * Save the output matrix using OMS as a sav file.
> OMS /SELECT TABLES
>  /IF COMMANDS = ["Factor Analysis"]
>      SUBTYPES = ["Reproduced Correlations"]
>  /DESTINATION FORMAT = SAV
>   OUTFILE = "C:\Program Files\SPSS\pomoc.sav".
> FACTOR  /VARIABLES bigband to rap  /ANALYSIS bigband to rap
>  /PRINT REPR EXTRACTION /rot=norot.
> omsend.
>
> * Purge the file.
> GET  FILE='C:\Program Files\SPSS\pomoc.sav' /drop Command_ Subtype_
> Label_ var2.
> sel if var1 = "Residual".
> save outf ='C:\Program Files\SPSS\pomoc.sav' /drop var1.
>
> * Transform all variables into one and plot its distribution.
> GET  FILE='C:\Program Files\SPSS\pomoc.sav'.
> VARSTOCASES /ID = id /MAKE resid FROM all /NULL = DROP.
> var lab resid "FA Residuals".
> EXAMINE  VARIABLES=resid
>   /PLOT BOXPLOT STEMLEAF.
>
> Greetings
>
> Jan
>
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of
> Susan Elgie
> Sent: Monday, August 22, 2005 9:55 PM
> To: SPSSX-L@LISTSERV.UGA.EDU
> Subject: Residual correlation graphs
>
> Hello!
> I am going to try again with my question, since there were no replies
> last time.  I need to run factor analyses from correlation matrices
> (quite a few times) and make stem-and-leaf plots of the residual
> correlations.
>
> I think this might be approached either computationally (e.g. by using
> the factor coefficients to score the original data and compute
> correlations and subtract them etc.) or by reading in the residual
> matrices which appear in the output.
>
> Either approach involves a formidable amount of computing for me; my
> knowledge of syntax is just the statistical part mostly.  If anyone
> could help me understand how to get started on either of these methods
> it would be great.  Or maybe I've got it all wrong; that would also be
> good to know.
>
> Many thanks,
>
> Susan Elgie
>
> -------------------
>
> Susan Elgie
> QQ Consulting
> Toronto, Canada
>
> qq.consulting@sympatico.ca
> selgie@sympatico.ca
>