Asymptotic expansions for the distribution of the sojourn time of a random walk on a half-axis (Q2446221)

Let \(X_{n}\), \(n\geq 1\), be a sequence of i.i.d. random variables, and let \(S_0=0\), \(S_{n}=X_1+\dots +X_{n}\), \(n\geq 1\), denote its associated random walk. Assume that the distribution of \(X_1\) has an absolutely continuous component, that \(\operatorname{E}X_1=0\), and that the Laplace transform \(\phi\) of \(X_1\) satisfies \(|\phi (\lambda)|<\infty\) for \(|\mathrm{Re}\lambda |\leq \beta\), \(\beta>0\). For \(b>0\), put \(T_{n}(b)=\sum_{k=1}^n I{S_{k}>b}\). The present work continues the author's studies [Sib. Math. J. 51, No. 1, 119--127 (2010); translation from Sib. Mat. Zh. 51, No. 1, 146-155 (2010; Zbl 1202.60067)] and [ibid. 51, No. 4, 621--638 (2010); translation from Sib. Mat. Zh. 51, No. 4, 785-804 (2010; Zbl 1217.60034)]. The main aim is to get asymptotic expansions of the probabilities \(\operatorname P(T_{n}(b_{n})=k)\) as \(n\to\infty\), where \(b_{n}/\log n\to\infty\) and \(b_{n}=o(n)\) as \(n\to\infty\).