1: \begin{abstract}

2: We study extremal statistics and return intervals in \emph{stationary}

3: long-range correlated sequences for which the underlying probability density

4: function is bounded and uniform.  The extremal statistics we consider

5: (e.g. maximum relative to minimum) are such that the reference point from

6: which the maximum is measured is itself a random quantity. We analytically

7: calculate the limiting distributions for independent and identically

8: distributed random variables, and use these as a reference point for

9: correlated cases. The distributions are different from that of the maximum

10: itself (i.e. a Weibull distribution), reflecting the fact that the

11: distribution of the reference point either dominates over or convolves with

12: the distribution of the maximum. The functional form of the limiting

13: distributions is unaffected by correlations, although the convergence is

14: slower. We show that our findings can be directly generalized to a wide class

15: of stochastic processes.  We also analyze return interval distributions, and

16: compare them to recent conjectures of their functional form.

17: \end{abstract}

18: