In this work we introduce two main models in extreme value theory (EVT) based on the theorems proposed by Fisher&Tippett 1928, Gnedenko 1943 and Balkema&de Haan 1974, Pickands 1975, respectively. As a crucial result in the classical EVT the former suggests the theoretical idea behind the block maxima model: by existence of weak convergence of appropriately centered and normalized maxima of iid rvs the only possible choice of non-degenerate limit distribution is a generalized extreme value (GEV) distribution. The latter provides mathematically solid foundations for the model for threshold exceedances and implies that the generalized Pareto distribution (GPD) is a natural model for excess distribution over a high threshold. In the real world extremal events often represent themselves through statistical data: Good levels of rivers, large insurance claims, large decreases in values of stock indices, stress losses on a portfolio of risky securities over a certain period of time, etc. Hence in this work we also introduce statistical methods for extremal events such as fitting a particular probabilistic model to real data, estimating the magnitude of stress event, etc
«
In this work we introduce two main models in extreme value theory (EVT) based on the theorems proposed by Fisher&Tippett; 1928, Gnedenko 1943 and Balkema&de; Haan 1974, Pickands 1975, respectively. As a crucial result in the classical EVT the former suggests the theoretical idea behind the block maxima model: by existence of weak convergence of appropriately centered and normalized maxima of iid rvs the only possible choice of non-degenerate limit distribution is a generalized extreme value (GEV) distribut...
»