Table 2. Estimated counts of families with an evolutionary bias towards vs. against ergatives, across methods, taxonomies and areas.
| Area | Bias | autotyp | glottolog | ||
|---|---|---|---|---|---|
| Binom./MCMC | ML | Binom./MCMC | ML | ||
| Africa | none | 4.17 | 1.26 | 3.27 | 3.27 |
| Africa | A | 34.83 | 28.74 | 41.73 | 41.73 |
| Africa | E | 0.00 | 0.00 | 0.00 | 0.00 |
| Eurasia | none | 1.29 | 8.49 | 7.30 | 7.32 |
| Eurasia | A | 28.71 | 24.12 | 23.69 | 23.67 |
| Eurasia | E | 0.00 | 3.39 | 3.01 | 3.01 |
| Pacific | none | 8.50 | 30.21 | 15.00 | 14.98 |
| Pacific | A | 24.11 | 14.02 | 21.14 | 21.23 |
| Pacific | E | 3.40 | 3.78 | 5.86 | 5.79 |
| South America | none | 12.00 | 12.01 | 12.47 | 12.48 |
| South America | A | 9.99 | 9.98 | 10.03 | 10.00 |
| South America | E | 2.01 | 2.01 | 2.50 | 2.52 |
| Rest of the Americas | none | 24.03 | 4.12 | 4.27 | 4.23 |
| Rest of the Americas | A | 18.76 | 34.88 | 35.73 | 35.77 |
| Rest of the Americas | E | 5.21 | 0.00 | 0.00 | 0.00 |
The table reports the means of 10,000 extrapolations. The standard errors of these are all smaller than.03. The extrapolations are based on large-family estimates of biases using set-based methods (with binomial tests) and tree-based methods (using MCMC and ML techniques, as described in the Methods section). MCMC-methods fully converged with binomial test results, resulting in the same estimates of counts. The two variants of ML-methods used here (fitDiscrete and BayesTraits) agreed among themselves but resulted in slightly different total estimates compared to the set-based and MCMC-based counts. Codes: A, against ergatives; E, towards ergatives.