Computational Modeling of Cognitive Processes with Bayesian Mixed Models in Julia
This project aims at writing an open-access book on cognitive statistical models in Julia.
Psychological and behavioural data that typically result from cognitive processes are often exhibiting characteristics that are not well captured by traditional statistical models. This issue has been simply ignored for a long time, with researchers using simple linear models without even thinking about whether they are appropriate, contributing to the replication crisis. Recent advances have underlined the need for statistical models that better reflect the data at hand.
Cognitive models are statistical models that best fit psychological data (e.g., reaction times, scales from surveys, ...) and can offer new insights by enabling inferences about the underlying cognitive processes that led to its generation.
Julia - the new cool kid on the scientific block - is a modern programming language with many benefits when compared with R or Python. Importantly, it is currently the only language in which we can fit all the cognitive models under a Bayesian framework using a unified interface like Turing and SequentialSamplingModels.
Unfortunately, cognitive models often involve distributions for which Frequentist estimations are not yet implemented, and usually contain a lot of parameters (due to the presence of random effects), which makes traditional algorithms fail to converge. Simply put, the Bayesian approach is the only one currently robust enough to fit these complex models.
As this is a fast-evolving field (both from the theoretical - with new models being proposed - and the technical side - with improvements to the packages and the algorithms), the book needs to be future-resilient and updatable by contributors to keep up with the latest best practices.
This project can only be achieved by a team, and I suspect no single person has currently all the skills and knowledge to cover all the content. We need many people who have strengths in various aspects, such as Julia/Turing, theory, writing, making plots etc. Most importantly, this project can serve as a way for us to learn more about this approach to psychological science.
If you are interested in the project, you can let us know by opening an issue or getting in touch.
See current WIP table of content.
- Fundamentals of Bayesian Modeling in Julia
- On Predictors
- Choices and Scales
- Reaction Times
- Individual Differences