Table 1. Model error rates.
As described in the results, we assessed each model’s prediction accuracy by calculating the mean absolute difference between the model prediction and the observed data across the bins in Figs 2–5. For each set of data (task + choice/RT) the most accurate model is indicated in bold. The asterisk in the first row indicates that the Bolton-Ockenfels model was actually fit to the choice data from Task 1, so it is not surprising that it provides the best fit.
| Error rates | aDDM | Fehr-Schmidt | Bolton-Ockenfels | Ornstein-Uhlenbeck |
|---|---|---|---|---|
| Task 1 choice | 2.9% | 4.2% | 1.7%* | 12.9% |
| Task 1 RT | 5.9% | 15.9% | 7.7% | 53.4% |
| Task 2 choice | 4.5% | 8.0% | 16.6% | 13.9% |
| Task 2 RT | 4.2% | 3.8% | 9.5% | 49.1% |
| Task 3 choice | 7.4% | 8.8% | 17.3% | 14.8% |
| Task 3 RT | 7.3% | 7.7% | 12.9% | 47.7% |
| Task 4 choice | 5.7% | 15.1% | 21.2% | 17.3% |
| Task 4 RT | 6.3% | 15.3% | 20.2% | 36.2% |
| Mean choice | 5.1% | 9.0% | 14.2% | 14.7% |
| Mean RT | 5.9% | 10.7% | 12.6% | 46.6% |
| Overall mean | 5.5% | 9.9% | 13.4% | 30.7% |