Classifying Behaviors in Risky Choices

This paper presents a nonparametric approach to classification of data from lottery experiments. Using very basic mathematical tools the paper endeavors to answer the questions: How to determine the “average” subject in a group? How to find a subject presenting the most similar behavior to a given one? How to detect outlier subject(s)? How to classify behaviors by their dissimilarity from the perfectly rational decision maker? How to rank subjects by risk attitudes? How to cluster subjects? This paper demonstrates that the answer to all of these questions may be found non-parametrically, without the use of any specific model.