This might be a simple question (I'm sure it is), but it's also my first here, so please go easy...on the other hand I'm possibly conflating a few points...

I'm helping a charity out, and have a bundle of result sets from various campaigns - individuals responding to an email message. I can't see the underlying data, but I can see the response broken down by Mosaic Codes (I can also get the Mosaic composition of who was targeted).

The trouble is that I have 239 different Mosaic Codes; and only about 50 sets of results.

I'm concerned that I'll get spurious results if I have so many possible explanatory variables. I could just run an analysis at a "higher" Mosaic level, where there are only c15 different categories, but they are keen to drill deeper.

There are clearly some relationships between different Mosaic codes - is this a case for dimension reduction?

  • $\begingroup$ Can you explain "Mosaic codes"? $\endgroup$ – kjetil b halvorsen Nov 25 '16 at 11:11
  • $\begingroup$ Categorical variables; in a hierarchy of 3 levels (11, 61, 239). They are compiled by a data agency to reflect geodemographic types. These are then appended to individuals by address lookups. So arguably, the obvious solution is to use the higher level, and run the 50 sets of results against those 11, which are in some ways a dimensional reduction of the 239. But I'm not sure if I'm thinking about this the right way. $\endgroup$ – N Mason Nov 25 '16 at 18:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.