This repository contains portfolio assignments, which were done during the Computational modeling for cognitive science course as part of Cognitive Science BA´s degree at Aarhus University
Teacher: Riccardo Fusaroli & Fabio Trecca (Eye-tracking workshop)
Instructor: Kenneth Christian Enevoldsen
2020 spring semester
In this course we start again from the foundations of statistical thinking: What is like to build a model? How do our expectations and previous knowledge shape and constrain models? How do we assess whether our models are any good and where they fail? We adopt a Bayesian framework implemented in Stan (R-packages rethinking and brms), since this approach makes it easier to build your own models and make your assumptions explicit. During the course we cover the generalized linear mixed effects model (GLMM) framework, but the course prepares you to go beyond this, via additional online materials and/or in the bachelor thesis.
File Portfolio_Exam.pdf contains a compilation of written reports solving the assignments listed below. Each portfolio assignment has a dedicated folder displaying the code solving the assignment (rmd and md files) in this repository.
| Assignment | Description |
|---|---|
| Portfolio 1: Eye-tracking data analysis | Preprocesing and analysing eye-tracking data collected in a lab during a Foraging experiment and Social engagement experiment |
| Portfolio 2: Evaluating the Cognitive science knowledge of Cognitive science teachers | Assessing rates from a binomial distribution, using the case of assessing our teachers’ knowledge of CogSci |
| Portfolio 3: Causal Inference | Exploring some issues related to multiple regressions (regressions with more than one predictor), and inferred (causal) relations between variable while utilizing Bayesian modeling. Diagnosing schizophrenia from multiple symptoms, utilizing causal graph method (Directed Acyclical Graph) to investigate the questions. |
| Portfolio 4: The role of priors and cumulative science | Running a Bayesian meta-analysis of pitch variability in ASD, based on previously published literature, analyzing pitch variability in ASD in two new studies using both a conservative and a meta-analytic prior, assessing the difference in model quality and estimates using the two priors. |
Knowledge:
- demonstate understanding of key weaknesses and strengths of both frequentist and Bayesian statistical methods
- demonstrate understanding of the basics of sampling and Bayesian inference
- demonstrate understanding of key distributions for Bayesian inference in cognitive science.
Skills:
- build and evaluate Bayesian models of hierarchically structured data
- compare models within the Bayesian framework
- build the appropriate Bayesian model given a research question and a data set
- communicate the assumptions, process and results of Bayesian data analysis.
Competences:
- plan and program data analysis projects relying on Bayesian modeling techniques.