Date:         Thu, 22 Oct 1998 16:25:11 GMT
Reply-To:     whiterr@MY-DEJANEWS.COM
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         whiterr@MY-DEJANEWS.COM
Organization: Deja News - The Leader in Internet Discussion
Subject:      Re: help with transforming correlation matrix

In article <7216EF1DF187D111A40500A0C9896EE5338F1A@ws11.kmv.com>,
  "Zhang, Jing" <jing.zhang@KMV.COM> wrote:
> Dear All,
>
> I am wondering if anyone has done this before:
>
> I have a SAS output data set containing correlation matrix from PROC
> CORR as following:
> _TYPE_   _NAME_     VAR1   VAR2 ..........  VAR100
> COR       VAR1               1      0.45 ...........  0.65
> .                                    0.45
> .
> .
> COR      VAR100           0.12         ............... 1
> I am trying to convert it to the following pair-wise correlation without
> the "1"s  and without the redundant correlation
>  in the following format:
>
> NAME1   NAME2      COR
> VAR1      VAR2        0.45
> .              .                .
> .              .               .
> .              .               .
>
> VAR99     VAR100    ...
>
> I would appreciate very much any suggestions you may have, many thanks!
>
> Jing Zhang
 
Something like this (untested) might do the job:
 
DATA CORROUT(KEEP=NAME1 NAME2 COR);
 SET CORR(DROP=_TYPE_ RENAME=(_NAME_=NAME1));
 ARRAY VAR[100] VAR1-VAR100;
  DO NUM=_N_+1 TO 100;
   COR=VAR[NUM];
   OUTPUT;
  END;
 
This assumes you want the "triangle" VAR1*(VAR2-VAR100), VAR2*(VAR3-VAR100),
VAR3*(VAR4-VAR100), ..., VAR98*(VAR99 VAR100), VAR99*VAR100.
 
Hope that helps,
    - Ray
 
-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own