Statistical inference in behavior analysis: Having my cake and eating it?

Abstract

Using simple, nonparametric statistical procedures can formalize the process of letting data speak for themselves, and can eliminate the gratuitous dismissal of deviant data from subjects or conditions. These procedures can act as useful discriminative stimuli, both for behavior analysts and for those from other areas of psychology who occasionally sample our journals. I also argue that changes in publication policies must change if behavior analysts are to accurately discriminate between real, reliable effects (hits) and false alarms.