Rates of convergence of some multivariate Markov chains with polynomial eigenfunctions (Q1024904)

Using spectral techniques, the authors provide a sharp nonasymptotic analysis of rates of convergence to stationarity for some multivariate discrete time Markov chains. All of them have multivariate orthogonal polynomials as eigenfunctions. The examples include the Moran model in population genetics and its variants in community ecology, the Dirichlet multinomial Gibbs sampler, a class of generalized Bernoulli-Laplace processes, a generalized Ehrenfest urn model, and the multivariate normal autoregressive process. The authors use what they call the chi-square distance to stationarity after \(\ell \) steps, starting at state \(x\in X\), that is defined as \[ \chi _{x}^{2}( \ell ) =\int_{X}\frac{( K^{\ell }( x,x^{\prime }) -m( x^{\prime }) ) ^{2}}{m( x^{\prime }) }\mu ( \mathrm{d}x^{\prime }). \] Here \(X\) is the state space of a general Markov process, \(K^{\ell }\) the \( \ell \)-step transition density and \(m\) the stationary density, both with respect to a \(\sigma \)-finite measure \(\mu \). The commonly used total variation distance \(d_{\mathrm{TV}}( K^{\ell }( x,\cdot ) ,m) =\frac{1}{2}\int_{x}| K^{\ell }( x,x^{\prime }) -m( x^{\prime }) | \mu ( \mathrm{d}x^{\prime }) ,\) yields a lower bound of \(\chi _{x}^{2}( \ell ) ,\) as \(4d_{\mathrm{TV}}^{2}( K^{\ell }( x,\cdot ) ,m) \leq \chi _{x}^{2}( \ell )\).