Top Two MLE Using Disjoint & Sliding Block Maxima
Poster, Bernoulli-IMS-Conference 2024, Ruhr Universität Bochum, Bochum
The research explores the estimation of Generalized Extreme Value (GEV) parameters using Top Two order statistics in both disjoint and sliding block maxima approaches, aiming to improve accuracy over traditional methods that only use maximum values. Results show that incorporating the Top Two significantly reduces mean squared error (MSE), especially in cases of small relative sample sizes, with sliding blocks further enhancing performance by increasing sample size. The study has implications for modeling extreme events, such as weather extremes, where capturing multiple significant values within a time frame can lead to better statistical estimates.