A shift in strategy or "error"? Strategy classification over multiple stochastic specifications

摘要: We present a classification methodology that jointly assign s to decision maker best-fitting strategy for set of choice data as well stochastic speci fication strategy. Our utilize normalized maximum likelihood model selection criterion compare multiple, possibly non-nested, specifications candidate strategies. In addition sing le with “error” specifications, we cons ider mixture i.e., strategies comprised probabilit y distribution over multiple this way, our approach generalizes the framework Brode r and Schiffer (2003a). apply an existing dataset find some makers are best fi t by single varying levels error, while others described using specification ov er