Table 4. Model comparison between a null model (one set of model-based and model-free regressors for both stimulation conditions) and more complex models that allow for an effect of tDCS on model-based control, model-free control, or both, which shows the null model is significantly more plausible than any of the models that allow for an effect of tDCS on behavioral control.
Model | No. of regressorsper subject | BIC | ΔBIC | AIC | ΔAIC | Bayes factor in favor of null model based on AIC |
null model | 13 | 18553 | 0 | 17752 | 0 | – |
separate model-freeregressors for Active and Sham | 16 | 18962 | 409 | 17796 | 44 | 1.3×1019 |
separate model-basedregressors for Active and Sham | 16 | 18947 | 394 | 17781 | 29 | 3.9×1012 |
full model | 19 | 19453 | 900 | 17852 | 100 | 2.7×1043 |