Using molecular dynamics simulations we determine the composition dependence of the self-diffusivity and transport diffusivity of a methane/ethane mixture at high pressure. We compute the transport diffusivity in two ways. First, the transport diffusivity is generated from the simulated self-diffusivities using an approximation known as the Darken equation. Second, the transport diffusivity is generated from the simulated phenomenological coefficients, based upon linear irreversible thermodynamics (LIT). We discuss the relative advantages of the two methods in terms of. (i) accuracy and (ii) computational demands of the approach. We find that the Darken equation gives values of the transport diffusivity within 6% of the more rigorous approach and is subject to substantially less statistical error with less computational effort. We find that the mean and standard deviation of the transport diffusivity obtained from linear irreversible thermodynamics are strong functions of the implementation of the infinite-time limit required in the evaluation. We suggest and implement an algorithm for statistically reliable transport diffusivities from molecular dynamics simulations. (C) 2004 Elsevier B.V. All rights reserved.