Power and sample size calculations for discrete bounded outcome scores