What happens if some pairs of options never occur together in a choice set? How can JMP help evaluate the design?
Feb 11, 2019 11:36 AM
When pairs of factor levels that never show up together in any choice set are always considered the best or worst by all participants, it could render the MaxDiff analysis impossible. For example, in this potato chip choice example, we found that some pairs of flavors (for example, Gyro and Dill Pickle) were not presented to any participants. In other words, they never occurred together in a choice set. If each participant considered them to be worse than all other flavors, then the MaxDiff model won’t run. This is simply because we won’t be able to rank these two flavors one over the other.
As with any experiment, it is important to evaluate the design carefully before implementing it. Use pairwise Incidence Matrix and Heatmap for each design generated. Then, create more designs to increase the number of pairings, if needed.