The composites industry is increasingly using molecular dynamics (MD) simulations to inform its materials development decisions. As a result, there is growing awareness that simulated predictions require quantitative assessments of their quality in order to routinely provide reliable and actionable information. In the following, we develop a suite of uncertainty quantification (UQ) tools designed to assess simulation-based estimates of the glass transition temperature T-g of polymer systems for aerospace applications. We consider contributions to this uncertainty arising from: (i) identification of asymptotic regimes in density versus temperature relations; (ii) fluctuations associated with limited time-averaging of dynamical noise; (iii) and finite-size effects associated with partial averaging over polymer-network configurations. We present a sequence of analyses by which we assess each of these contributions and quantify their net effect on estimates of T-g. Importantly, these methods suggest more efficient workflows by indicating when multiple small simulations can be combined to yield estimates with uncertainties comparable to larger, more expensive simulations. We expect that related approaches will, in the future, be applicable to other physical quantities of interest as well as to a broader class of computational tools. (C) 2016 Elsevier Ltd. All rights reserved.