Clustering non-quantitative as well as quantitative variables

[Nina Bhanap <NBhanap@Zefer.com>]

Hi there

I have a set of variables, some of which are binary, some are numeric and some are qualitative. I want to run a cluster analysis on these variables, but not sure the best way to represent them quantitatively.

1) So for example, if I have {online; ATM; branch; mail; call centre}
If I represent them as {1;2;3;4;5}, then this is not really accurate, because "call centre" will be plotted further away than "online" and this is not necessarily correct.

How then do I convert these variables into sensible quantitative numbers?

2) If I have a qualitative variable with one of two states {happy; sad}, should I convert these to {1;2}? Does this make sense?

3) Can I combine variables of type in 1) and 2) as well as numerical variables like "bank balance"?

Thanks in advance - Nina

:::.:::..::::.:.:::.:::.::::

//nina.bhanap
//business.strategy
//zefer.san francisco
//vox.415.762.3487
//fax.415.247.9929

:::::.:::::::.:::.:::.:::::