Table 3.
Model comparison of reinforcement learning model fits to subject data.
| Model | FP | −LL | AIC | PEP | EF |
|---|---|---|---|---|---|
| TRAINING | |||||
| Standard | 3 | 13543.8 | 27705.7 | 0.007 | 0.239 |
| Learning bias | 5 | 13298.2 | 27626.4 | 5.081e-05 | 0.003 |
| Bayes HT | 4 | 13402.9 | 27629.8 | 0.339 | 0.362 |
| Decision bias | 4 | 13284.2 | 27392.4 | 0.654 | 0.397 |
| TEST | |||||
| Standard | 3 | 4755.4 | 10128.8 | 3.816e-16 | 0.164 |
| Learning bias | 5 | 4362.8 | 9755.7 | 3.816e-16 | 0.002 |
| Bayes HT | 4 | 4748.4 | 10320.8 | 3.816e-16 | 0.015 |
| Decision bias | 4 | 4304.6 | 9433.3 | 1.000 | 0.818 |
FP: number of free parameters; −LL: negative log-likelihood; AIC: Akaike information criteria; PEP: protected exceedance probability, the probability that a given model is the most frequent in the population, above and beyond chance; EF: estimated model frequency, the frequency of the model in the population as estimated by the Bayesian random-effects analysis.