feature Importance in MBC

Question

Hi,

We have conducted an MBC exercise with 15 attributes (each tested across 5 price levels).

For analysis - we have created 15 models to predict the choice of the attributes using HB regression.

We wanted to understand if we can arrive/calculate the feature importance of the 15 attributes using the HB price utilities?

Regards,
Surabhi

Answer 1

Surabhi,

I don't know what value importance calculations give you over and above what you'll get from running simulations.  

To be consistent with how folks estimate (also not obviously useful) importances in conjoint analysis  I suppose you could compute the range in utilities across the 15 features and the 15 ranges in utilities across each of the features' price points.  Then any one of these ranges divided by the sum of the 16 ranges would be its importance.  

Again, I do not know how this will be valuable.