Let’s talk about the distinction between Main Effects and Interaction Effects. The store color&lighting experiment in Chapter 9 gives a nice overview of these ideas, but interaction effects can be tricky, so they’re worth exploring more.
Let’s consider the following experiment. Suppose that Miller 64 is considering changing the look of the bottle. They are looking at a slimmer, taller, bottle shape – which, they believe would be more consistent with the “skinny” image of the brand. In order to test the new design, they conduct a laboratory experiment. The subjects are randomly selected from a consumer panel, and then randomly assigned to one of the 2 experimental conditions:
- “Redesigned” condition: the subjects are asked to make a choice among 5 different beer brands (Miller 64 and the closest 4 competitor brands). All the choices are presented in their actual bottles, with Miller 64 being shown in the new, slimmer bottle.
- “Traditional” condition: the subjects have the exact same task, with the difference that Miller 64 is shown in its existing, regular type of bottle.
The results show no difference in the proportion of subjects choosing Miller 64 under the 2 conditions. The company concludes that the expenses incurred with redesigning the bottle are not a good idea, since the design does not have any effect on choice.
To begin with, what are Main Effects and Interaction Effects? Why should we care about interactions?
Can you think of any potential interaction effects that could have occurred in this experiment? Do you think that Miller’s conclusion is necessarily the right one, given this research study?