Abstract
The SMEMAX Transformation was tested on samples generated from log-Pearson Type III and Gumbel Extreme Value Type 1 distributions. Although the transformation results in a distribution with a skew near zero, it is not a true normal distribution. The coefficients of kurtosis for the transformed distributions are always less than three and when the parent distribution approaches a reverse-J shape, the transformed distribution is bimodal. When the skew coefficient of the logarithms of events is -0.5 or greater for samples from the log-Pearson Type III distribution, the SMEMAX procedure underestimates the magnitude of events with return periods of 50 yr and greater. Similarly, SMEMAX estimates of rare events from the Gumbel Extreme Value Type 1 distribution are always too low. Based on comparisons made in this study, it is likely that application of SMEMAX procedures to any skewed distribution will result in biased estimates of the magnitude of rate events and if the skew is positive the estimates will be too low. © ASCE.
Recommended Citation
J. A. Westphal, "Smemax? Caution!," Journal of Hydraulic Engineering, vol. 110, no. 3, pp. 209 - 218, American Society of Civil Engineers, Jan 1984.
The definitive version is available at https://doi.org/10.1061/(ASCE)0733-9429(1984)110:3(209)
Department(s)
Civil, Architectural and Environmental Engineering
International Standard Serial Number (ISSN)
1943-7900; 0733-9429
Document Type
Article - Journal
Document Version
Citation
File Type
text
Language(s)
English
Rights
© 2023 American Society of Civil Engineers, All rights reserved.
Publication Date
01 Jan 1984