My question concerns a particular finding from a randomized, placebo-controlled within-group cross-over study. I have an interaction effect of relationship duration by treatment condition on ratings of romantic partner attraction. Specifically, under Placebo condition, the long-term relationship duration group rated romantic partners very low on attraction, but short-term relationship duration group rated partners very high on attraction. Whereas, under the treatment condition (in this case intranasal oxytocin), both groups moved closer to the mean, which resulted in a significant condition by relationship duration interaction. What I am wondering is whether RTM is applicable in this case given the cross-over design. In other words, OT was time 2 for half the sample and Time 1 for the other half. So, if you are looking at the effect from Time 1 to Time 2, half the sample moved away from the mean and half toward the mean. I have tested the groups independently by procedural order (i.e., whether they received treatment visit 1 or 2), and the effect is the same.
Related to your query, there is some useful well studied literature that may be of your special interest:
Article Regression toward the mean in 2 × 2 crossover designs with b...
Article On the Proper Use of the Crossover Design in Clinical Trials...
Article On the Proper Use of the Crossover Design in Clinical Trials...
Hello Harry,
I didn't read anything in your description of the study to suggest that the regression effect should be a concern. You never selected or assigned cases on the basis of extreme scores (the most obvious circumstance for regression to the mean).
Further, your wise choice to counterbalance the condition sequence across participants (and subsequent check that sequence made no difference) strengthens the argument that the treatment group shows less difference (as a function of relationship length) than does the placebo condition.
Good luck with your work!
Thank you for the quick and helpful responses. Muhammed, I have sent off for the first article - looks promising. David, thank you for looking over my methods and providing your expert opinion - certainly gives me more confidence in the findings and in how I discuss this potential limitation in the paper.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics