Cognitive Models
Computational Modeling of Cognitive Processes with Bayesian Mixed Models in Julia
Preface
Why Cognitive Models?
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.
Why Julia?
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.
Why Bayesian?
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.