Conditioned limit theorems for products of positive random matrices

Abstract: Inspired by a paper of I. Grama, E. Le Page and M. Peign\'e, we consider a sequence $(g_n){n \geq 1}$ of i.i.d. random $d\times d$-matrices with non-negative entries and study the fluctuations of the process $(\log \vert g_n\cdots g_1\cdot x\vert ){n \geq 1}$ for any non-zero vector $x$ in $\mathbb R^d$ with non-negative coordinates. Our method involves approximating this process by a martingale and studying harmonic functions for its restriction to the upper half line. Under certain conditions, the probability for this process to stay in the upper half real line up to time $n$ decreases as $c \over \sqrt n$ for some positive constant $c$.