Rotations in Product Tests and Surveys
Ennis, D. M., Rousseau, B., and Ennis, J. M. (2014). IFPress, 17(1) 3-4.
In this report we consider how to balance item positions, item sequences, and the spread of these sequences within rotations. The term “sequence” will be used to mean a pair of items occurring one after the other. The term “item” refers to a product, concept, or a question in a survey. Methods to account for sequential effects in product tests, or events and behaviors in survey research, play a role in product testing and survey design. Generally, interest centers on balancing the effect of one item on another to minimize bias and reduce test variance. There have been a number of approaches to addressing the design of tests so that sequential effects are accounted for and these include random presentation orders, the use of Latin Square designs, and computer searching through a large number of designs to meet a balancing criterion. In product testing over multiple days, which becomes necessary when large number of products are evaluated, sequences within days are much more important than between days and it is useful to consider how designs for this situation can be constructed.
Figure 1. Number of times that the sequences occur in the Position five position orders and how they are clustered around 10.