Max-semistable extreme value laws for autoregressive processes with Cantor-like marginals

This paper considers a family of autoregressive processes with marginal distributions resembling the Cantor function. It is shown that the marginal distribution is in the domain of attraction of a max-semistable distribution. The main result is that the extreme value law for the autoregressive process is obtained by including an extremal index in the law for an i.i.d.\ process with the same marginal distribution. Connections with extremes in deterministic dynamical systems and the relevance of max-semistable distributions in that context are also pointed out.