On the multidimensional stochastic equation \(Y_{n+1}=A_{n} Y_{n}+B_{n}\) (Q704266)

The aim of this note is to extend the existing results of the behaviour at infinity of the tail for the stationary solution of the multi-dimensional linear auto-regressive process described by the stochastic equation: \[ Y_{n+1} = A_nY_n + B_n, \quad n \in \mathbb N, \;Y_n \in\mathbb R^d,\;d \geq 2,\tag{SE} \] where \((A_n, B_n)\) is a sequence of i.i.d. (independent and identically distributed) random variables, \(A_n\) belongs to the linear group of invertible square matrices of size \(d\), and \(B_n\) is a vector in \(\mathbb R^d\). The existing results for the solution of (SE) were obtained for coefficients being either non-negative matrices or non-singular matrices. The authors succeed to provide weaker conditions on the coefficient sequence \((A_n, B_n)\) in (SE), namely a property called i-p-e condition (irreducibility-proximality-expanding), under which the solution to (SE) has a polynomial queue behaviour at infinity. The key point of the proof relies on deriving a certain renewal theorem for functionals of the Markov chain involved with the matrix coefficient \(A_n\).