Asymptotic \(\Gamma\)-distribution for stochastic difference equations (Q1185781)

Let \(F_ 0\subset F_ 1\subset\dots\) be an increasing sequence of \(\sigma\)-fields on some probability space and let \(X_ 0,X_ 1,\dots\) be real-valued random variables which satisfy the following assumptions: \(X_ n\geq0\), \(X_ n\) is \(F_ n\)-measurable; there are functions \(g(x)\) and \(\sigma^ 2(x)\), \(x\geq 0\), and random variables \(\xi_ 1,\xi_ 2,\dots\) such that a.s. \[ X_{n+1}=X_ n+g(X_ n)+\xi_{n+1},\quad E(\xi_{n+1}\mid F_ n)=0,\quad E(\xi^ 2_{n+1}\mid F_ n)=\sigma^ 2(X_ n). \] Assume also that there is a \(C>0\) and a \(\delta>2\) such that for all \(n\geq 0\) a.s. \[ E(|\xi_{n+1}|^ \delta\mid F_ n)\leq C\sigma(X_ n)^ \delta. \] The author is interested in the behaviour of \(X_ n\) on the event \(C_ \infty=\{X_ n\to\infty\), as \(n\to\infty\}\). More specifically he investigates the conditional distribution of \(X_ n\), given \(C_ \infty\). The main result states that if the magnitudes of \(g(x)\) and \(\sigma^ 2(x)\) are properly balanced at infinity, then \(X_ n\) is asymptotically \(\Gamma\)-distributed in a suitable scale. This result contains several known theorems.