A Comparison of L-Moments with Method of Momentsby Bingzhang Lin, Maryland Univ at Coll Park, Silver Spring, United States,
John L. Vogel, Maryland Univ at Coll Park, Silver Spring, United States,
Abstract: A comparison of the Conventional Moments Method (CMM) with the L-Moments Method (LMM) for selecting a distribution is made. Five distributions GEV (Generalized Extreme Value), LNO (Lognormal), PE3 (Pearson Type III), GLO (Generalized Logistic) and GPA (Generalized Pareto) were used as the parent models for a regional frequency analysis of 228 daily and 122 hourly rainfall stations in Pennsylvania. Scatter diagrams of skewness and kurtosis and L-skewness and L-kurtosis were used to define the goodness-of-fit by applying four criteria. A Real-Data Check (RDC) was used to confirm the results. The CMM indicated that the PE3 was the optimum distribution, while the LMM indicated that the GEV was the optimum distribution. The RDC indicated that the GEV is the most suitable distribution and the LMM is the more desirable method for parameter estimation. The results show that the GEV II is more suitable for modeling the extreme events, while the GEV I or Gumbel is the worst.
Subject Headings: Comparative studies | Statistics | Moment (mechanics) | Data processing | Parameters (statistics) | Logistics | Rainfall frequency | Mathematical models | Frequency analysis | Pennsylvania | North America | United States
Services: Buy this book/Buy this article
Return to search