On record indices and record times

Abstract

In this paper conditions are characterized under which the record indices of a sequence of independent, non-identically distributed random variables (with not necessarily continuous distribution functions) are independent, or, equivalently, the corresponding record times form a Markov chain with conditionally invariant transition probabilities (CITP chain). In particular, some former results of Nevzorov are simplified and generalized. The correlation structure of record indices in the general case is also studied. © 1995.