Uncertainty in MD-Derived Diffusion Coefficients Depends on Analysis Protocol, Not Just Simulation Data

Bonkowski and De Souza [Sol. Stat. Ionics 429, 116967 (2025)] provide a guide for performing molecular dynamics simulations of ion transport, including methods for estimating diffusion coefficients and their uncertainties from mean-squared displacement (MSD) data. The discussion of uncertainty in estimated diffusion coefficients presents the statistical properties of specific analysis protocols—ordinary least squares regression with particular data processing choices and other, statistically equivalent, approaches—without explicitly stating the scope of applicability for these results. Without explicit scope statements, the guide presents these results as universal characteristics of diffusion coefficient estimation; i.e., uncertainty in D* is determined only by the input simulation data. Here, we clarify that for diffusion coefficients estimated by linear regression of MSD data, the uncertainty depends not only on the input simulation data, but also on the choice of statistical estimator (OLS, WLS, GLS) and data processing decisions (fitting window extent, time-averaging). By clarifying this dependence on analysis protocol, we hope to help researchers avoid incorrect uncertainty estimates and encourage the adoption of statistically efficient methods for analyzing molecular dynamics simulations.