Quantification of extrapolation performances of extreme quantile estimators
Résumé
The analysis of the extreme data helps regulate the norms of installation of energy sites such as nuclear plants and hydraulic dams in accordance with the adaptation to floods, storms, droughts and meteorological aggression coming with climate change. A specific model for extreme values needs to be designed, and confidence intervals need to be established to capture a quantile of a selected probability. It is important to refine the confidence intervals to help in decision-making. Standard confidence intervals are established in common libraries in R language, but the results obtained on simulated data and real data cannot properly fit the theoretical objectives. The goal is to present new ways to construct confidence intervals on large quantiles using extreme-value theory in the special case of heavy-tailed distributions and study the efficiency of the newly-built intervals.