| Date: | Thu, 25 Sep 2003 21:25:46 +1000 |
|---|
| Reply-To: | Doug <dougie_cam@HOTMAIL.COM> |
|---|
| Sender: | "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU> |
|---|
| From: | Doug <dougie_cam@HOTMAIL.COM> |
|---|
| Organization: | Global Crossing Internet |
|---|
| Subject: | cluster analysis in Everitt and Der |
|---|
Hi,
I have been working through some examples in Everitt and Der - 'A Handbook
of Statistical Analysis using SAS' and looking at their cluster analysis of
prestige data from SDS in chapter 9 (p115-127). In this chapter they use
the observation number as a variable in the analysis. I'm no expert on
cluster analysis but I don't believe that using the observation number in
the analysis is valid. However if the observation number is removed, then a
large part of their analysis appears to be invalid. Has anyone had a look
at this chapter and found the same problem. I'm sure that they are only
using this as an example but I just can't understand why an arbitrary value
such as observation number would be useful.
Doug
|
