ABSTRACT
A confidence interval for the difference between two populations’ centers can be especially informative when the dependent variable is measured on the same familiar scale across the studies in an area. However, often dependent variables are abstract and are measured indirectly using relatively unfamiliar measures. For example, consider research that compares Treatments a and b for depression, a variable that is more abstract and more problematic to measure directly than would be the case with the familiar dependent variable measures that were listed in chapter 2. Although depression is very real to the person who suffers from it, there is no single, direct way for a researcher to define and measure it as one could do for the familiar scales. There are many tests of depression available to and used by researchers, as is true for many other variables outside of the physical and biomedical sciences. (In such cases the presumed underlying variable is called the latent variable, and the test that is believed to be measuring this dependent variable validly is called the measure of the dependent variable.) For example, suppose that confidence limits of 5 and 10 points mean difference in Beck Depression Inventory (BDI) scores between depressed groups that were given Treatment a or b are reported. Such a finding would be less familiar and less informative (except perhaps to specialists) than would be a report of confidence limits of 5 and 10 lb difference in mean weights in our earlier example of research on weight gain from treatments for anorexia.
