I'd like to know if anyone out here has used Rosenthal and Rubin's Binomial Effect Size Display (BESD). I'm finding it hard to interpret correlation coefficients calculated on continuous data using the BESD, and I could use some help. Rosenthal and Rubin suggest understanding the r as a change in success rate. I could calculate a correlation coefficient on the use of, let's say, behavior therapy for cocaine use, and my r could be r = .60. I'm not sure what that really means in terms of success rate. The BESD says it would mean that it's equivalent to a change in success rate from .20 to .80. Does that mean that without the behavioral intervention only 20% of people would change for the better, and that with it, 80% change for the better? Or, does it mean that the intervention improves outcome 60% over what it would be without the intervention? And, if I calculated the data using a within groups design, how would I know if there's a 60% improvement without the intervention? I'd appreciate any help possible.

Bill Gottdiener New York