An extension of almost sure central limit theorem for order statistics (Q626289)

Let \(M_n\) and \(m_n\) be the first and the second maximum of an i.i.d. sequence \(X_1,\dots, X_n\). Assume that, for some \(a_n>0\), \(b_n\) and a nondegenerate CDF \(G\), \((M_b-b_n)/a_n\) converges weakly to \(G\). Then, for any bounded Lipschitz function \(f\) on \(\mathbb R^2\), \[ \lim_{n\to\infty} {1\over D_n}\sum_{k=2}^n f \left({M_k-b_k\over a_k},{m_k-b_k\over a_k}\right)=\int fdH \text{ a.s.}, \] where \(H(x,y)=G(y)(\log G(x)-\log G(y)+1)\), \(D_n=\sum_{k=1}^n d_k\), \(d_k\) is a number sequence satisfying some mild conditions, e.g., \(d_n\sim (\log n)^\varepsilon/n\) for \(\varepsilon>0\).