Modeling a Large Number of Agents by Types: Models as Large Random Decomposable Structures

摘要: This paper introduces methods, based on decomposable random combinatorial analysis, to model a large number of interacting agents. also discusses largely ignored possibility in the mainstream economic literature that hitherto unknown types agents may enter models at some future time. We apply notion holding times, and introduce results one- two-parameter inductive methods Ewens, Pitman Zabell literature. More specifically, we use exchangeable partitions finite set produce simple rule sucession, is, expressions for probabilties entries by new or known types, conditional observed data. Then Ewens equilibrium distriution sizes clusters is introduced, its examine market behavior sketched, especially when few are dominant. suggest approaches this times relevant agent-based simulations because can be used randomly select “act” first.