Multiple DVs

Hey all,

I've been tasked with determining the "best" IV (various demographics) for predicting overall purchase behavior, as determined by behavior across multiple product categories (multiple DVs).

Which of these demographics is the best predictor of individuals overall purchasing routine?

Age (13-17, 18-24, 25-34, etc)
Lifestage (Teens, Single/no kids, married/no kids, married/kids, etc)
Cohort (Millennials, Gen X, Boomers)

Would need to determine if age, lifestage or cohort is the "best" way to segment consumers (ie best way to predict overall behavior).

I can't reduce the DV's to a single DV, so i'm wondering the best approach to this. DV's are expressed quantitatively (amount purchased of given category).

I suppose maybe MANOVA, but i wonder about having multiple Categorical variable breaks in my IVs is an issue. Every example i've seen has only two breaks for their IV when using MANOVA.

Any thoughts!?


TS Contributor
Hi justcallmeerik,

Why can't you reduce the DV's to a single DV? I ask you this because in my opinion, the best alternative would be doing exactly that. The way I see it, you have a common latent variable problem. Purchasing behavior is some sort of latent variable that you cannot directly measure but you can estimate it using the information regarding the purchase information available. Using Factor Analysis to reduce the problem and then a multiple regression analysis would be enough.

There are multivariate regression techniques and MANOVA could also be used, but I'd consider the first one as the most elegant solution.