Date:         Fri, 11 Sep 1998 12:17:06 +0200
Reply-To:     Marcel Dijkstra <marcel@EUREKA.NL>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         Marcel Dijkstra <marcel@EUREKA.NL>
Organization: Universiteit Maastricht
Subject:      Re: Aggregate & Median scores
Content-Type: text/plain; charset=us-ascii; x-mac-type="54455854";
              x-mac-creator="4D4F5353"

Thank you Richard for the reply.
 
I figuered out the same answer to the problem. However, it will not solve
the problem. I simplified the the question a little bit in order to make
things not to complex. I have said that I always have 75 subjects. However,
in the final analysis I have to split up these 75 cases in four groups. Not
knowing in advance how big these groups will be. If I would rank the
variable say in 100 parts, and I would use the 50th part as my median, I
frequently get a non existing value back (the 50th value does not alway
exist).
I will try David Marso's sollution to the problem.
 
Marcel Dijkstra
Junior Statician
Neuropsychology & Psychobiology
University of Maastricht, The Netherlands
 
 
Richard F Ulrich wrote:
 
> In syntax, you might solve your problem with:
> RANK   TIME by SUBJ into TIMERANK
>
> and then you would select for timerank= 38, for the median of 75 scores.
>
> If you have any Missing or Ties, you might want to score as percentiles,
> and then (for one example), select as 50%ile  or less, then aggregate
> by Subject to get the MAX of those scores under 50%ile.
>
> I have never performed this myself, but it seems simple enough that it
> ought to work.
>
> --
> Rich Ulrich, biostatistician                wpilib+@pitt.edu
> http://www.pitt.edu/~wpilib/index.html   Univ. of Pittsburgh