On Wed, 10 Apr 2013, rfro...@jbu.edu went:
I think the problem is in the use of the term "power" as synonymous with sample size.
I agree. The Buttons et al. paper never seemed to say anything about within-subject designs (even though a lot of fMRI studies probably use such designs). And that's one place where power and sample size decouple. If you tested five different doses of a drug, including placebo, in a Latin-square design using just five participants (each receiving all five doses), you'd have plenty of power--more than you'd have if you used a between-groups design with 25 participants (each receiving one of the five doses). But with your five-person Latin square, you'd be more vulnerable to nonreplicability due to idiosyncrasies of your sample, which is what Buttons et al. are concerned about. At least, that's my thinking. Someone set me straight if need be. --David Epstein da...@neverdave.com --- You are currently subscribed to tips as: arch...@jab.org. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5&n=T&l=tips&o=24930 or send a blank email to leave-24930-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu
