Luca, You mean weighting VARIABLES, not CASES (the latter being the more usual application of weighting). When applying clustering procedures, the unit of measurement of variables is irrelevant (all are standardized internally to a zero mean and unit standard deviation), so that, say, multiplying your two priority variables by some factor would have no effect. One possibility is forgetting about clustering altogether, and run a factor analysis on the seven variables. The risk is that FA may not give higher weight to the two "chosen" variables but to some other ones. Data are stubborn sometimes. Another possibility is classifying the cases only by those two priority variables, into a small number of classes (this can be done by cluster analysis or some other procedure), and then refining the classification by means of the other 5. I guess the number of groups may thus grow very large, though. At the moment I do not have more precise ideas about this. Perhaps others may come up with something better. Hector

-----Mensaje original----- De: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] En nombre de Luca Meyer Enviado el: Friday, February 10, 2006 10:57 AM Para: SPSSX-L@LISTSERV.UGA.EDU Asunto: Weighted cluster analysis

Hello,

I have been asked to run cluster analysis on a database to identify omogeneous groups of cases. I have also been asked to weight a couple of variable more than the other 7 so that groups will depend more heavily on variability on these 2 variables. What would be my best choice for procedures? Should I look into AnswerTree?

Thanks, Luca

PS: I only have a license for SPSS Base and Ctables, I do not have other SPSS products

Mr. Luca MEYER Survey research, data analysis & more: http://lucameyer.blogspot.com/ Tel: +390122854456 - Fax: +391782232575 - Mobile: + 393394950021