Extreme values statistics for Markov chains via the (pseudo-) regenerative method (Q626299)

The authors apply the regenerative method and the Nummelin technique to the estimation of extremal properties of a time series \(f(X_1),f(X_2),\dots,\) where \(X_1\), \(X_2\),\dots is a time-homogeneous \(\psi\)-irreducible Harris recurrent Markov chain. Estimates for the tail index and the extremal index are considered based on maxima over the approximate pseudo-renewal cycles of the chain. Applications to AR(1) and ARCH(1) models are discussed. Results of simulations are presented.