The logarithmic average of sample extremes is asymptotically normal. (Q1879507): Difference between revisions

Let \(X_1, X_2,\dots \) be independent and identically distributed random variables, \[ \xi _k = \left.\left (\max _{1\leq i \leq k} X_i - a(k)\right )\right/b(k) \quad\text{and}\quad S(n) = \sum _{1\leq i\leq n} \frac {1}{i}h(\xi _i), \] where \(h\) is a function of bounded variation having compact support and \(a(k), b(k)\) are centering and norming sequences. It is supposed that there is a non-degenerate distribution function \(H\) such that \(\lim _{n\to \infty } P(\xi _n \leq t) = H(t). \) A strong approximation theorem for \(S(n)\) is proved. The weak convergence of \(S(n^t), 0\leq t \leq 1,\) and the law of the iterated logarithm then follow as consequences. The asymptotic variance of \(S(n)\) is derived in dependence on the domain of attraction of the distribution function \(H.\)