Abstract:
|
Extreme Value Theory (EVT) deals with the extreme deviations from the median probability distribution and is used to study rare but extreme events. When considering the use of EVT to model data where extremes exist, one must consider whether extreme events are stationary or non-stationary. This study examines few methodological issues involving non-stationary events by investigating the assumption of a series of independently and identically distributed data through time (stationary). In particular, the spot prices for West Texas Intermediate (WTI)crude oil data from 1986 to 2016 have been modeled. Considering that there are many factors that cause fluctuation in crude oil prices such as supply and demand, natural disasters and various world crises, hypothesis testing have been used to determine whether the data is stationary or non-stationary. With the conclusion that the data follow a GEV distribution, model selection is used to see which one of the three families within GEV would best model the crude oil data. After selecting the best fit model, the bootstrapping method has been used to construct confidence levels for extreme values in non-stationary conditions.
|