Date: Wed, 24 Jul 2002 12:51:24 +0200
Reply-To: Stefan Kniffka <Stefan.Kniffka@t-online.de>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Stefan Kniffka <Stefan.Kniffka@t-online.de>
Subject: Coding variables in clusteranalysis with spss
Content-Type: text/plain; charset="iso-8859-1"
Hi members of the list,
for my exam I have to do a clusteranalysis with SPSS. By coding the categorial and ordinal variables i used the methodes from Backhaus and Steinhausen/langer and coded the categorial variables with n categories in n binar variables. The ordinal variables with n steps i coded in n-1 binar variables.
Here is my problem:
If am using Tanimoto (or other methodes, which only countes item, which are existing at both objects), the result by coding 00 little, 10 medium, 11 big for two objects with 00 will be 0, but it should be 1. If I am using Simpe Matching (or other methodes, which countes item, which are existing or not existing at both objects), I will get the correct result for the example above, but not for the previous categorial variables: Example 100 green colour; 010 red colour, 001 black colour; the result for 100 and 010 is 1/3, but it should be 0.
Shure, I can use two different methodes and than aggregate; but this is not possible in SPSS.
So, who can help ??
Thank you for your help!