Box-Cox Transform and Its Application in Flood Frequency Analysis

This paper studies three methods for estimating the Box-Cox parameter used to transform data to normality. The proposed methods are based on finding the parameter values that optimizing the Anderson-Darling, ShiproWilk and relative mean square error tests for normality. The performance of the these methods combined with normal distribution are evaluated and compared with the most applied distributions for flood frequency analysis, namely generalized extreme value (GEV). The performances of these distributions are tested using the annual maximum flood data of Johor stations. The results show that the normal distribution after transformed data to normality performed as well as the GEV distribution and they are recommended for further analysis.