Tail calculus with remainder, applications to tail expansions for infinite order moving averages, randomly stopped sums, and related topics (Q2488464)

Let \(X_i\), \(i\in Z\), be i.i.d. copies of a r.v. with regularly varying survival function \(\overline F(x)=\mathbf{P}\{X_i\geq x\}\). Denote \(\langle a,X\rangle=\sum_i c_iX_i\). The asymptotic expansions for the survival function of \(\langle a,X\rangle\) are investigated of the form \[ \mathbf{P}\{\langle a,X\rangle \geq x\} =\sum_{0\leq i\leq m}p_{\infty,i}t^{-\alpha-i}+o(t^{-\alpha-m}), \] where \(p_{\infty,i}\) can be represented in terms of analogous coefficients for \(\overline F(t)\) and \(F(-t-0)\). E.g., if \(X_i\) have a Student-t distribution with the parameter \(\alpha\) and \(Y_i\) is stationary with \(Y_i=aY_{i-1}+X_i\), \(| a| <1\), then \[ \mathbf{P}\{Y_i>t\} ={K_\alpha \alpha^{(\alpha-1)/2}\over 1-| a| ^\alpha} {1\over z^\alpha}+{p_2\over t^{\alpha+2}}+O\left({1\over t^{\alpha+4}}\right), \] \[ p_2=K_\alpha\alpha^{(\alpha+1)/2}{\alpha(\alpha+1)\over \alpha-2} \left( {1\over 2(1-| a| ^2)(1-| a| ^\alpha)} -{\alpha\over (\alpha+2)(1-| a| ^{\alpha+2})} \right), \] \(K_\alpha=(\sqrt{\alpha} B(1/2,\alpha/2))^{-1}\).