A psychologist got into a verbal exchange with a colleague about two modes of behavior therapy for patients with photonumerophobia.* He likes a gradual increase in exposure to the noxious stimulus whereas his buddy throws his patients into the deep end and then gives lots of support. To resolve the issue, they do a study in which the DV is galvanic skin response, an index of anxiety, measured while subjects are delivering a research paper. After the lecture, subjects are then asked to recall how anxious they were before treatment by using a rating scale, to be used as a covariable. The covariate was significant (p <0.05), but no treatment differences were found.
*Fear that their fear of numbers will come to light. Question. Who wins the bet?
Answer. No one. Because the covariate was measured after treatment, people who responded well may rate their initial anxiety lower than those who didn't respond. Then, if one method worked better, the final anxiety in this group would be lower, but so would the initial anxiety, leading to an underestimate of the treatment effect.
Was this article helpful?