Maxima of recurrent random sequences (Q1851496)

The paper deals with the random recurrent sequences \(\{X_n\}\), \(n\geq 0\), of the form \( X_n = a_nX_{n-1}+b_n\), where \(\{a_n\}\) and \(\{b_n\}\), \(n\geq 1\), are sequences of independent nonnegative random values with distributions \(A\) and \(B\), respectively, \( X_0\geq 0\); \(\mathbf Ma_n = \alpha<1\), \(\mathbf Mb_n <\infty\), \(\mathbf MX_0 <\infty\); \( X_0\), \(\{a_n\}\) and \(\{b_n\}\) are independent in totality. The author studies the behaviour of maxima \(M_n = \max\{X_0,\dots,X_n\}\) as \(n\to\infty\). For \(\bar B(x) = \exp\{-cx+o(x)\}\), \(c>0\), \(x\to+\infty\) and \(A(1-0)=1\) it is proved that \(\mathbf P(M_n\leq u_n)\to e^{-\tau}\), \(n\to\infty\), where \(u_n\) are such that \(\bar\Psi(u_n)\sim \tau/ n\), \(n\to\infty\), and \(\Psi\) is a stationary limiting distribution of \(\{X_n\}\).