Date:         Wed, 4 Feb 1998 12:39:33 -0600
Reply-To:     swheeler@MOORMAN.COM
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         swheeler@MOORMAN.COM
Organization: Deja News Posting Service
Subject:      Re: grouping obs

Sorry if this is a repeat.
 
Try using PROC FASTCLUS with MAXCLUSTER= # of stores
                             Radius = 0
 
Here is an example:
 
/*
    Start with 1 observation per store.
    If necessary, Use PROC TRANSPOSE to form the data set
*/
 
Data Stores;
  Input Store X1-X8;
Cards;
1 1 2 3 4 5 6 7 8
2 1 2 3 4 5 6 7 8
3 2 2 3 4 5 6 7 8
4 1 3 4 5 6 7 8 9
5 2 2 3 4 5 6 7 8
6 9 8 7 6 5 4 3 2
7 9 8 7 6 5 4 3 2
8 9 8 7 6 5 4 3 2
9 9 8 7 6 5 4 3 2
;
/* Use PROC FASTCLUS to cluster the stores.
     SET MAXCLUSTERS = # of Stores
         RADIUS = 0
            you can increase RADIUS if you do not need strict equality
            see SAS STAT Manual for definition of distance
*/
 
PROC FASTCLUS DATA=Stores MAXCLUSTERS=9 radius=0 OUT=clusters;
  VAR x1-x8;
  ID Store;
RUN;
 
/* Print output data step
    you are on your own from here
*/
PROC SORT DATA=Clusters;
  BY CLUSTER;
RUN;
 
PROC PRINT data=CLUSTERS;
RUN;
 
Stan Wheeler
 
In article <19980204060601.BAA27123@ladder02.news.aol.com>,
  emwalczak@aol.com (EMWalczak) wrote:
>
> Thanks for the ideas!!!  But I don't think I explained the problem very well.
> I need to look at store#1 52 weeks of data and see if  store#2 has exactly
the
> same prices over those 52 weeks.
> but if store #2 doesnt maybe store#3 looks like store #1 and Store #2 looks
> like store #4 etc..... I have 800+ stores that I want to look at over 52
weeks,
> does that make more sense?   I was thinking proc cluster might do the trick,
> but have no clue on that proc!
>
> I need to group these 800+ stores in to different groups based on their 52
week
> spending trends.
>
> Any help would be great, I hate the idea of going thru this 2 stores at a
time!
> Thanks,
> Ellen
 
-------------------==== Posted via Deja News ====-----------------------
      http://www.dejanews.com/     Search, Read, Post to Usenet