Penultimate limiting forms in extreme value theory (Q2266535)

Let \(X_ 1,...,X_ n\) be i.i.d. random variables with distribution function F and let \(M_ n\) denote their maximum. \textit{B. Gnedenko}, Ann. Math. 44, 423-453 (1943), showed that the properly normalized sequence \(M_ n\) can only converge to one of the following distribution functions: \(\Lambda (x)=\exp (-\exp (-x)),\) \(x\in {\mathbb{R}}\); \(\Phi_{\alpha}(x)>\exp (-x^{-\alpha}),\) \(x\geq 0\); or \(\psi_{\alpha}(x)=\exp (-(-x)^{\alpha}),\) \(x\leq 0\). In the first case, the author shows for a broad class of functions that an approximation by distribution functions of the other cases gives a much better rate.