INTERCOMPARISON OF PARAMETER ESTIMATION METHODS IN EXTREMAL WAVE ANALYSIS

Abstract

Based on statistical analysis of extreme wave height data generated with a Monte-Carlo simulation technique for the prescribed parent probability distributions, a preferable method for the parameter estimation was determined for each of 8 distributions. It is also verified that a jackknife method is applicable to the correction of bias and the estimation of variance irrespective of parameter estimation method in most parent distributions, and that the information matrix methods inherent to the maximum likelihood method give generally satisfactory results in the estimation of variance of return wave height for samples of size greater than around 50.