Date:         Tue, 4 Apr 2000 20:04:14 -0400
Reply-To:     Raynald Levesque <rlevesque@VIDEOTRON.CA>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Raynald Levesque <rlevesque@VIDEOTRON.CA>
Subject:      Re: Finding Median for subset of Records
Content-type: text/plain; charset=iso-8859-1

You may find the following useful (It was posted by David Marso to the
newsgroup comp.soft-sys.stat.spss some time ago.)

How to aggregate with the median:

GET FILE='C:\Program Files\SPSS\1991 U.S. General Social Survey.sav'.
* A LITTLE TRICK TO BE SURE RMV HAS SOMETHING TO RMV TO * .
SORT CASES BY occcat80.
MATCH FILES FILE * / BY OCCCAT80 / FIRST=TOP.
DO IF TOP.
XSAVE OUTFILE "FILL.SAV" / KEEP occcat80.
END IF.
EXECUTE.
ADD FILES / FILE "FILL.SAV" / FILE * /IN=MED_DEV/ BY OCCCAT80.

SPLIT FILE BY OCCCAT80.
PRESERVE .

* HERE IS THE MAIN TRICK NOTE A LIST OF VARIABLES MAY BE USED WITH RMV * .
SET RESULTS=OFF.
RMV MED_AGE=MEDIAN(AGE,ALL).

** In THIS CASE I "match" the median and difference all appropriate cases
**.
** YOU COULD JUST SELECT IF NOT(MED_AGE) to get your "Aggregate" File: .

IF MED_DEV MED_AGE=LAG(MED_AGE).
COMPUTE MED_DEV=AGE-MED_AGE.
EXE.

*CLEAN UP * .
SELECT IF NOT TOP.
MATCH FILES FILE * / DROP TOP.
EXECUTE.
RESTORE .

HTH

Raynald Levesque rlevesque@videotron.ca
----- Original Message -----
From: Dan Heuman <DANH@POLK.CA>
Newsgroups: bit.listserv.spssx-l
To: <SPSSX-L@LISTSERV.UGA.EDU>
Sent: Tuesday, April 04, 2000 12:36 PM
Subject: Finding Median for subset of Records


> I basically need a way of doing an Aggregate command and return the median
> for each break variable.  Aggregate doesn't have median as an option.  I
> could do a split file of the break variable and run frequencies but that
> won't save the results to a data table.  Am I missing something?  Is there
a
> way to do this?
>
> Thanks,
>
> Danny Heuman
> danh@polk.ca