Date:         Wed, 16 Jul 1997 11:01:27 +0100
Reply-To:     Jeremy Miles <J.N.V.Miles@DERBY.AC.UK>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         Jeremy Miles <J.N.V.Miles@DERBY.AC.UK>
Subject:      Re: correl. of 1 variable with recd. type for pairs of records
Comments: To: Hank Becker <hjbecker@UCI.EDU>
In-Reply-To:  <33C286F1.5A69@uci.edu>
Content-Type: text/plain; charset="us-ascii"

At 10:29 08/07/97 -0800, Hank Becker wrote:
>I have a data file where the cases are in pairs of records (much like
>one would have with separate pre- and post-test files merged into a
>single file). I want to compute simple correlations for each variable
>between that variable and the "record type" (i.e., for matched pairs of
>records constituting a single case).
>
 
 
If I have understood correctly, you can create a new variable using the LAG
command.  This will create a new variable, shifted down one case.
 
So:
 
COMPUTE newvar = lag(oldvar) .
 
 
Will turn
 
Oldvar  Newvar
1               [missing[
2               1
3               2
4               3
5               4
6               5
                6
 
 
Now remove all of the odd numbered cases using:
 
SELECT IF(mod($casenum, 2) = 0).
 
Will give:
 
 
Oldvar  Newvar
2               1
3               2
4               3
6               5
 
 
 
 
===============================================================
Jeremy Miles, AVRU, Derby University, DE3 5GX Fax: 01332 622287
Tel: 01332 622222 x. 2146               j.n.v.miles@derby.ac.uk
======================No Space for a Joke======================