Goodness-of-fit Criteria for the Adams and Jefferson Rounding Methods and their Limiting Laws

摘要

Multiplier methods are used to round probabilities on finitely many categories to rational proportions. Focusing on the classical methods of Adams and Jefferson, we investigate goodness-of-fit criteria for this rounding process. Assuming that the given probabilities are uniformly distributed, we derive the limiting laws of the criteria, first when the rounding accuracy increases, and then when the number of categories grows large.