| Clearly, how much scores change with therapy is not just a function of the sensitivity of the measure but also of how much real change there was over the period of the therapy. Unless it’s clear that the measures are tuned to rather different aspects of change, we’d expect to see improvement in therapy correlating fairly strongly across change measures and this is one key bit of evidence that a change measure is fit for purpose. It’s one component of the validity of a change measure. |
Details #
Sensitivity to change can be measured in different ways but probably the most common are to report the mean change, ideally with its 95% confidence interval (CI) but another common approach is to report the effect size of the change (again ideally with the CI, though this is despicably rare!) There is a complication that reports that give the effect size often say they are using Cohen’s d as their measure of effect size without saying which of Cohen’s d values they are using: the issue is that change is a “paired” or “within person” measure and the Cohen’s d that is most often used for effect size is for comparison between groups. More on that here later!
Try also … #
Effect size
Confidence interval
Cohen’s d
Paired tests / within person tests
Chapters #
Chapter 3.
Online support #
I hope to put up some simulations illustrating the Cohen’s d2 and dz and an app allowing people to put in change data and get a full set of change analyses.
