Finite sample behaviour of the mixed moment estimator in dependent frameworks

In this paper, via Monte Carlo techniques and for dependent structures, like the max-autoregressive processes and the m-dependent processes, we explore the behavior of a recently introduced extreme value index estimator, the mixed moment estimator. The dependent stationary sequences considered provide a wide spectrum of dependency, with an extremal index ranging from a value close to one (as happens in identicallly distributed settings, where exceedances of high thresholds appear isolated) to any value smaller than one, a situation in which exceedances of high levels appear in clusters of a mean size approximately equal to the reciprocal of that extremal index.