Running regression for each cluster


New Member
Is this common or reasonable? Run a k-means cluster analysis and then do a separate logistic regression for each cluster?

For simplicity, let's say this is a marketing scenario... we have 3 clusters and I want to determine an individual person's probability of responding to a direct mail piece. So the idea is that since each cluster has it's own regression model, we'll get better results overall, as opposed to having only one model for the whole group that may be diluted.

Thx! Appreciate the input and any other direction you can provide.