As we can see on table 4, the computational time to obtain the pattern statistics for all simple words of a given length are quite similar with both approaches. One exception: the Bernoulli case (m = 0) where SR are roughly 10 times faster than FMCI computations. This is due to the fact that the Bernoulli case uses simpler recurrences than the ones used for the true Markovian cases (m ≥ 1). Similar simplifications in the DFA structures can reduce the computational time of FMCI approach in the independent case but they have not been implemented here (as their use is often marginal).