SELECTION OF PARAMETER-ESTIMATION METHOD FOR LP3 DISTRIBUTION

The quantile prediction acc;racy of the log-Pearson type III (LP3) dis tribution depends largely on the accuracy of the parameter-estimation method used. The performance of a parameter-estimation method, on the other hand, depends on both the individual population chosen from the LP3 family and the sample size. In this study Monte Carlo experiments were conducted to evaluate four parameter-estimation methods that are frequently used in hydrological analysis. The four methods tested are the method of indirect moments (MMI), the method of mixed moments (MIX ), the method of direct moments (MMD), and a modification of MMI using optimization techniques (MMO). A quantile ratio index (QRI) was devis ed to identify the limits (sample size and LP3 population subset) with in which each of these methods will perform best. This study suggests that when QRI less than or equal to 1.14, MMI or MMO should be used fo r sample size N less than or equal to 30, MIX for 30 < N < 100, and an y of the four methods for N greater than or equal to 100. When QRI > 1 .14, MMO is recommended for N less than or equal to 30, MIX for 30 < N < 100, and MIX, MMO, or MMI for N greater than or equal to 100. An ap plication procedure was also developed and successfully applied to 10 randomly selected sites in Louisiana.