TABLES
Table 1. Comparison of the four tested models, ranked by relative predictive accuracy (Expected Log-Posterior Density, elpd_diff), and associated uncertainty (se_diff). Model 4 is “best”, as per Vehtari et al. (2017) and (Sivula et al., 2020). Model 1 corresponds to the baseline model, to which an extra pathway was added for each of models 2-4.