I am interested in general approach/specific examples where Floquet multipliers for linear system of ode with periodic coefficients are determined, at least approximately, by perturbation arguments. Specifically, I have in mind a linear system, parametrized by epsilon, such that: 1. for epsilon=0 the system is defined by a constant-coefficient diagonalizable matrix; 2. for epsilon>0, the system has periodic coefficients. How may I "follow" the motion of the multiplier as epsilon increases (at least for small epsilon). More than that: what may happen in case of singular perturbations?