Biased transformed kernel estimator of extreme quantiles

• Catalina Bolancé ^[1] ; Ramon Alemany ^[1]
  1. [1] Universitat de Barcelona

     Universitat de Barcelona

     Barcelona, España

     
• Localización: Contributions to risk analysis: risk 2018 / coord. por José María Sarabia Alegría, Faustino Prieto Mendoza, Montserrat Guillén Estany, 2018, ISBN 978-84-9844-683-8, págs. 79-86
• Idioma: inglés
• Texto completo no disponible (Saber más ...)
• Resumen

  • When the data are right skewed, the classical kernel estimator (CKE) does not smooth the right tail of the probability and cumulative distribution functions. For estimating probabilities and conditional probabilities associated with the extreme quantiles of a random variable, we propose an estimator that combines a parametric distribution assumption with a kernel estimator. To compare it to the CKE, we carry out a simulation study that shows how our proposal is better when the distribution is heavy-tailed. Finally, we apply the methodology to the estimation of extreme quantiles in the age-at-death distribution of the Spanish population.