2001 Volume 45 Pages 187-192
This paper exmamines the usefulness of a probability distribution with lower-and upper-bounds for small samples with a size of 20 to 30. Small samples for more water periods and less water periods are extracted from a long-term (10, 000) series of lognormal variates generated by the Monte Carlo technique. The accuracy of T-year event estimates is assessed when the probability distribution with upper-bound is applied to these small samples. The results here recommend that for such samples that include large events the incorporation of upper-bound gives better T-year event estimates. It avoids overestimation of T-year quantiles. However, the use of the probability distribution with no upper bound is recommended for such small samples that only small events are included in the sample. The use of the probability distribution with upper bound for such samples tends to underestimate the T-year events.
