In this paper, we use probabilistic methods to analyze learning effects in a behavioral experiment on the newsvendor model. We argue why we should believe that suggested orders follow a multinomial logit distribution, and ...