Resampling based Bias Correction for Estimation in Generalized Extreme Value Distribution

The generalized extreme value distribution is modeled to approximate the extreme individualities of a process. Our goal is to evaluate the performance among the estimation techniques of maximum likelihood estimator(MLE), penalized MLE and L-moment estimator with taking in account of bias. We try to improve the performance of likelihood method and it is found that the penalized maximum likelihood estimator(PMLE) with beta prior is the best. We introduce resampling(Bootstrap and Jackknife) based bias correction approaches. Based on simulation study, it is observed that the amount of bias obtained by Jackknife method is larger than Bootstrap. PMLE is recommended to use instead of MLE and L-moment estimator for negative case of the shape parameter. All of our erudition move towards to the Jeju wind speed data.