Regression results.

For each model (Model), we present the effect of a given predictor variable on the log-odds of cooperative versus competitive games—positive values indicate variables that are associated with an increased relative frequency of cooperative goal structures. The top block of estimates are from models omitting phylogenetic controls (Phylo.), and the bottom block of estimates are from models that control for phylogenetic signal in the outcome using phylogenetically constrained random effects. The top block corresponds with the “Basic” plots in Fig 3 and the bottom block with the “Phylogenetic controls” plots in Fig 3. In the univariate models, the column Predictor gives the effect of the indicated variable. In the multivariate models, the column Predictor gives the effect of the indicated variable and the column Interdependence land gives the effect of interdependence in land-based hunting. Intervals are central 90% posterior credible intervals. D gives the relative WAIC (Watanabe-Akaike Information Criterion) value compared to the best model, and W gives the unit normalized WAIC weight across both phylogenetic and non-phylogenetic models.