Check parameter recovery

[GidonFrischkorn]

Once, we have all models in place and the distribution functions are programmed, we should run some basic parameter recovery studies. These should give a rough intuition about how much data is needed for the different models and how choices, such as the choice_rule and links affect parameter recovery. We could even consider exploring model misspecification, for example simulating data using identity links and evaluating how choosing different links such as log would affect the results.

• Parameter recovery for different number of data: 50, 100, 250 responses per subject per condition
• Parameter recovery for different choice_rule and links
• Evaluate model misspecification with incorrect links or choice_rules

[GidonFrischkorn]

This is something that we could add in the Appendix of the tutorial paper. Likely, reviewers would be interested in this anyway and it would be a nice complementation to the simulations that Jan Göttmann did together with Klaus and me.