Title: A comparative study on robust estimators of the extremal index Authors:  M Cristina Miranda - University of Aveiro (Portugal) [presenting]
Manuela Souto de Miranda - University of Aveiro (Portugal)
Ivette Gomes - FCiencias.ID, Universidade de Lisboa and CEAUL (Portugal)
Abstract: The extremal index is a statistical parameter that emerges in the study of clusters of extreme observations in stationary sequences, like exceedances of specific levels. Clusters of extreme values typically tend to arise when there is no independence in the observed data. Under some convenient conditions, the maximum limit distribution of such sequences has a distribution of the extreme value type, but it is affected by the value of the extremal index, $\theta$, which varies from 0 to 1. The accuracy of the estimated values of $\theta$ might be relevant in practical situations since $\theta$ is related to the average clusters dimension. Robust versions of $\theta$ most popular estimators are compared and their performance with simulations is analysed. The study considers two different models suitable for data series with a high number of zeros. The results indicate the advantages of using the robust versions in the presence of contaminated data.