I am working with collaborators who are conducting an education research study to see if there is a difference in learning outcomes between two learning experiences, one with a technology tool (let’s call it Tech) and without (let’s call it Read).

The researchers designed the study to be a crossover design, but I am not sure that it fits this definition, and as a result, I am struggling with the best way to analyze the data. The researchers split a class into two groups, one group studied topic A with Tech and the other group studied topic A with Read. Then, the groups were switched and they studied topic B with the other method. Pre- and post-tests for topics A and B were given after the learning sessions for each topic.

Since the topics, and thus the content of the pre-/ post-tests are different, should this be considered a crossover trial? Or is it just two separate experiments?

I think the crossover was done to make the experience similar between the two groups for human research ethics reasons, but this doesn’t mean that statistically the study is a crossover design similar to how a drug trial would be conducted (which uses just one drug).

I hope I am clearly describing what was done. Thank you for your help.


Yes, this fits the criteria for a crossover design. You could analyze this as a mixed model, with test scores nested by student. Even if the primary focus is on the effect of the tool (Tech or Read) on test scores, you may also want to include topic (A or B) as a predictor, so that you are estimating the effect of the tool on test scores, over and above the effect of the topic.

You can find a helpful discussion of nested effects here: https://stats.stackexchange.com/a/228814/154048


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.