Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

CBC: conditional prohibition of attribute levels based on respondent profile


we are about to conduct a CBC study with the following peculiar specs/needs:

there is a specific combination of attribute levels that should not be shown for respondent from certain states because the "product" would simply not exists at all in those states --- it would really not make sense for those respondents to see these attribute levels combined altogether.
So we have to hide them. It's no problem to do this with a bit of javascript by detecting the presence of those levels altogether in a cbc card.
I have two questions:

1- since this prohibited combination will not occur at every iteration of the exercise, there will be some iterations onto which we'll see 3 cards and some other iterations (most of them, actually) onto which we'll see 4 cards. Will this create a bias from the respondent's point of view -- the fact of being exposed to a variable number of cards between each iterations?

2- for this particular profile for which this "concept" will be hidden, the levels composing the latter will necessarily have more chances to end up with lower scores since they will be exposed less often and so, in a way, data for these levels will be less accurate than for the other ones. So the question: how are we to treat/interpret this during analysis?

VoilĂ . Many thanks in advance.

asked Feb 15 by sebas (235 points)

1 Answer

+1 vote
Best answer

This solution may cause more problems than it solves.  In cases like this I usually make two different CBCs, one with the prohibition and one without it.  Then I route respondents using skip logic into the state-appropriate CBC.  

Then on the back end I export each of my two experiments as a single .csv file, combine the two .csv files and then analyze with a single CBC experiment using CBC-HB.
answered Feb 15 by Keith Chrzan Platinum Sawtooth Software, Inc. (53,875 points)
selected Feb 19 by sebas