The importance of attribute interactions in the conjoint choice design and modelling Jie Yu, Roselin

1
The importance of attribute interactions in the
conjoint choice design and modellingJie Yu,
Roselinde Kessels, Peter Goos and Martina
VandebroekDepartment of Applied Economics,
Katholieke Universiteit Leuven, Belgium
Department of Mathematics, Universiteit
Antwerpen, Belgium
Problem Description Within the field of conjoint
analysis, there seems to be some doubt whether or
not the predictive performance of the model with
interactions would be better than that of the
main effects model. In real applications,
researchers might not know whether there are
attribute interactions. Therefore it is valuable
to find a proper design and a proper estimation
model that, when combined together, is most
robust against misspecifications of the
underlying model.
How to find it??

• Applications
• - Commercial applications
• interactions between price and brand
• - Gene technology for food production
• interactions between gene technology
• and price

• Questions and solutions

Question Which model should I use in the design
and estimation stages if I do not have any
information about the existence of interaction
effects?
Reason The predictive performance of the
interaction model is robust to the true
situations with different interaction effects
sizes and the true situations without interaction
effects, while the main effects model is not
robust
Answer Choosing interaction-effect-design in the
design stage and interaction-model in the
estimation stage is recommended
Answer The predictive performance tends to
increase
Question What happens if I increase the number
of alternatives within the choice sets?