I'm dealing with this big dataset which has:

  • 1 categorical variable with 90 levels that represent some sort of "geographical area"

  • 3 continuous variables

What I'm trying to do is to "aggregate" the levels of the categorical variable, to a maximum of 10.

I don't know exactly which technique I should use to accomplish this, for example whether I should use factor analysis or some clustering (unsupervised)

  • 2
    $\begingroup$ How about creating groups based on geographical regions, such as Northeast, Southwest, etc., that are more meaningful from the business perspective? $\endgroup$ – Vishal Apr 25 '16 at 17:58
  • $\begingroup$ Yes, but I also want to see if there is some analogy between areas in different regions, in order to investigate why! $\endgroup$ – user5609462 Apr 25 '16 at 19:00
  • $\begingroup$ What do you mean by "clustering (unsupervised)"? $\endgroup$ – Marquis de Carabas Apr 26 '16 at 21:50
  • $\begingroup$ Have a look at: stats.stackexchange.com/questions/227125/… $\endgroup$ – kjetil b halvorsen May 16 '17 at 23:43