Weak convergence of a stochastic numerical method

A stochastic numerical method has weak order $p$ when its error in expected sufficiently smooth observables is $O(\Delta t^p)$. Weak convergence measures approximation of distributions rather than individual sample paths.