Using the prior predictive in modeling cognition.

Psychologists are often faced with a plethora of models that all try to explain the same psychological behavior. In light of this proliferation of models, it is crucial to identify the model that best approximates the empirical data. To reach this objective of model selection, mathematical psychologists examine how well the data and model agree, but also how flexible the model is (model complexity). This research project will investigate how the prior predictive distribution can be used to measure model complexity and to perform model selection. The role of the prior predictive models for building models and designing of experiments will also be examined. Two major hurdles must be taken: developing meaningful and acceptable parameter priors and developing meaningful and acceptable data priors. In addition to developing a measure of complexity and a method for model selection based on the prior predictive is an important goal is to compare the new measure and the new method with previously proposed measures and methods.

Keywords:Data prior, Generalizability, Model selection, Informative prior, Bayesian statistics, Theory testing

Disciplines:Psychological methods, Mathematical and quantitative methods, General pedagogical and educational sciences, Social theory and sociological methods, Political theory and methodology, Animal experimental and comparative psychology, Applied psychology, Human experimental psychology, Biological and physiological psychology