Variational formulas for asymptotic variance of general discrete-time Markov chains (Q2108480)

The authors study the asymptotic variance of a discrete-time irreducible ergodic Markov chain with transition probability \(P\). It plays an important role in functional central limit theorems. They use the fact that, via the functional central limit theorem, asymptotic variance can be studied in terms of the Poisson equation of the form \((I-P)x = f\), where \(f\) is a given mean-zero square-integrable (with respect to the stationary distribution of the chain) function, and the solution \(x\) is a square-integrable (with respect to the stationary distribution of the chain) function. Assuming that the spectral radius of \(P\) is strictly less than one (which ensures the existence of the solution of the Poisson equation), they provide formulae for the asymptotic variance in terms of the bilinear form associated with \(I-P\); both in the general case and in the reversible chain case. They provide several applications of their results, with some emphasis on the non-reversible case.