Extremal index estimation for a weakly dependent stationary sequence (Q1317263)

Let \(\{\xi_ i\}\) be a strictly stationary sequence of random variables with a continuous marginal distribution \(F\). We say that the extremal index of \(\{\xi_ i\}\) is equal to \(\theta (\in [0,1])\) if \[ \lim_{n \to \infty} P \left( n \Bigl( 1-F \biggl( \max_{1 \leq i \leq n} \theta_ i \biggr) \Bigr) \geq x \right) = e^{-\theta x} (x>0). \] Under weak dependence, the author considers the estimation and statistical significance of the extremal index \(\theta\). For the first problem, he defines an estimator of the extremal index \(\theta(>0)\) by \[ \widehat \theta_ n (x) = \sum^ n_{i = 1} I \Bigl( \max_{i + 1 \leq j \leq i + r - 1} \xi_ j \leq x < \xi_ i \left. \Bigr) \right / \sum^ n_{i = 1} I(\xi_ i > x) \] and shows that, for suitably chosen \(r\), \(\widehat \theta_ n(x)\) is weakly consistent for a wide range of \(x\). For the second problem, the author shows that the distribution of the sample maximum can be uniformly approximated given the extremal index and the marginal distribution. For \(m\)-dependent sequences the same problems are discussed in detail.