Comparison of theories for payoff disbursement of coalition values (Q1084030)

A procedure is proposed for competitively testing coalition models of payoff disbursement which make point, line, or area predictions. By allowing the prediction set of a model to expand uniformly in the outcome space, the procedure generates for each model a support function which relates the cumulative proportion of observed payoff vectors within the expanded set of the model's predictions to the relative size of this set. And by comparing to each other the support functions of any two models while equating the relative size of their expanded prediction sets, the procedure offsets the advantage possessed by the less specific model prescribing a larger or more dispersed prediction set. The procedure is then employed to competitively test ten coalition models in two different characteristic function 3-person game experiments. Dominance relations among models are shown to depend on the power relations among players.