Grouping categories and not grouping observations with cluster analysis

[carlos viviano]

Greetings. How can I make a dendrogram group categories of a nominal variable and not the observations. In my database, the observations are "individuals", but I don't want to see how the "individuals" are grouped (there are many) but I want to see how the categories of the nominal variable "provinces" are grouped. Note: each province consists of a number of "individuals".

[Nick Cox]

You could formalize this as an analysis of variance and what follows, but the usefulness of that depends a lot on how many many provinces you have. If there are say 196 provinces I would try something different. I might want to collapse so that there were 196 observations to feed into a cluster analysis, or perhaps better something like principal component analysis. The collapse need not be just to means.

[carlos viviano]

You could formalize this as an analysis of variance and what follows, but the usefulness of that depends a lot on how many many provinces you have. If there are say 196 provinces I would try something different. I might want to collapse so that there were 196 observations to feed into a cluster analysis, or perhaps better something like principal component analysis. The collapse need not be just to means.

Thank you so much. My case is like your second example (many provinces), so I'll try collapsing the base.