A Further Investigation of Selten's Measure of Predictive Success

There are two basic ways of assessing the goodness of fit of theories to data - one based on stochastic theory (for example, the maximised likelihood in some form) and one based on deterministic theory, for example, Selten's Measure of Predictive Success. This paper explores the second of these and presents an application of Selten's Measure to the problem of comparing and ranking various theories of decision making under risk. The paper uses an experiment which was specifically designed to provide insight into the usefulness of this measure. Specifically the questions in the experiment were chosen to give a high degree of discrimination between the various theories being ranked by Selten's measure. However, in common with a previous application, it is found that the measure appears to fail because it has no mechanism for differentiating between observations inconsistent with the theories. This seems to be an inherent failing of a measure based on deterministic theory.