Table 3.
Results of Bayesian t-tests probing the hypothesis that model parameters and the interaction term of reward × state as a raw data indicator for the degree of model-based behavior is lower in participants with alcohol-dependent fathers as compared to participants without positive family history.
| Bayesian Independent Samples T-Test | ||
|---|---|---|
| B01 | error (%) | |
| Interaction score raw data | 6.691 | ~1.596e−4 |
| β1 | 1.447 | ~1.751e−4 |
| β2 | 1.838 | ~1.651e−4 |
| α1 | 11.172 | ~4.376e−6 |
| α2 | 4.290 | ~1.448e−4 |
| λ | 5.395 | ~1.489e−4 |
| ω | 5.933 | ~1.528e−4 |
| ρ | 7.781 | ~1.727e−4 |
In the table, we show evidence in favor of the null hypothesis. By convention, a Bayes factor > 3 indicates evidence in favor of the hypothesis (Kass and Raftery, 1995). The Bayes Factor is interpretable as an Odds Ratio: for example, it is more than six times more likely that the data occurred under the null hypothesis that interaction of reward and state does not differ between groups than under the alternative hypothesis of a difference between groups.