Fitting and Comparison of different Behavioral Models , i.e. Rescorla Wagner, Noisy Win-Stay-Lose-Shift, Epsilon Greedy, etc.
Analysis of behavioral data requires a good understanding of task designing and functionalities of the human body, especially circuits or target regions of the brain that we aim to study. With the lack of a good understanding, fitting and comprising of models are blinded and just a few measurements are available to apply in choosing the top model; however, no helpful explanation could be proposed for why the model works good or even be able to detect possible errors. In the procedure of analysis, there are some important steps like, designing a model based on the task, interpretability of the model, simulations (esp. parameter recovery), model validation, and model recovery that are skipped due to the lack of information.
- Random Model
- Rescorla Wagner
- Noisy Win-Stay-Lose-Shift
- Rescorla Wagner + choice kernel
- Epsilon Greedy
- AIC
- BIC
Epsilon Greedy Model has low variance for its parameters and because of that, it seems to be a better model in comparison to Rescorla Wagner and its extended model.
You could find more detailed information in the Report file