Coefficient accuracy comparison.

We report the accuracy of inferred selection coefficients β for three diffusion-based methods using the simulated data described in Sec. 3.1. We consider two metrics: root mean squared error (RMSE; top) and mean absolute error (MAE; bottom). In both cases the metric is normalized such that the value of the metric for β = 0 is equal to unity. For example, the RMSE is normalized by ||β*||₂, where β* are the true effects. We do not include a comparison to PyR₀, since it utilizes a somewhat different likelihood, making direct comparison subtle. See Sec. 3.2 for discussion.