Abstract
A Bayesian decision rule about the number of different values of a discrete randomvariable is numerically tested by Monte Carlo experiments. Results are exhibited showing the percentage of correct decisions as well as the robustness of the rule.