Experiment design: some options

One of Terracotta’s strengths lies in its ability to enable a variety of different experiment designs. This article details some, but not remotely all, of them.

Important: An exposure set (in Terracotta) is not the same as a period (in typical repeated-measures design parlance).

Example: Let’s say you’re going to use Terracotta to administer 6 assignments in your class, and you want one group of students to get the control version for the first 3 assignments, and another group to get the experimental version, and then at the midpoint, you’ll want those conditions to switch.

Some users assume that such an experiment should be set up this way:

...but Terracotta was designed to function differently. 

In the example above, there are 6 periods, but only 2 exposure sets. To achieve this design, here’s the proper configuration:

It’s important to note that Terracotta doesn’t control the timing of the assignments (so there’s no need to specify when the crossover occurs in Terracotta). The instructor configures the order of assignments in Canvas by setting open and close dates for each assignment. In our example, there’s only one crossover.

Another option:

What if the instructor wanted to have 5 crossovers, with students switching back and forth between experiment conditions sequentially: