Date:         Thu, 14 Jul 2005 02:36:18 -0400
Reply-To:     Richard Ristow <wrristow@MINDSPRING.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Richard Ristow <wrristow@MINDSPRING.COM>
Subject:      Re: Cluster Analysis: more questions
In-Reply-To:  <200507132227.j6DMRBfD031614@listserv.cc.uga.edu>
Content-Type: text/plain; charset="us-ascii"; format=flowed

At 06:27 PM 7/13/2005, Talbot Michael Katz wrote:

>SAS has three main clustering procedures, CLUSTER, FASTCLUS, and
>MODECLUS.  All three will accept your raw "coordinate" data; CLUSTER
>and MODECLUS also accept distance data.  FASTCLUS and MODECLUS will
>compute their own Euclidean distances based on
>your coordinate data (and it will treat latitude and longitude as if
>they are x and y coordinates, which they're not)  If all your data is
>reasonably localized -- say within the Northeast corridor, or the
>Southwest, or even the continental United States -- you might be able
>to get away with Euclidean distances, even based directly on latitude
>and longitude.

Over a small area, longitude and latitude work approximately as
Euclidean coordinages, but NOT of equal length. The distance
represented by a minute of latitude is constant; the distant
represented by a minute of longitude is the same constant times the
cosine of the latitude. Around 41 degrees latitude (north or south), a
longitude degree is about 0.75 as long as a latitude degree; that
correction applies no matter how small the distances are.