Pre-conference Events

BCCCD25 is offering pre-conference Tutorials

Places are limited, so early registration is recommended to secure your spot.

Fee: EUR 40/person

T01: Bayesian models of science learning in Python
Organizer: Lucas Lörch, DIPF Leibniz Institute for Research and Information in Education

Thursday, 9 January 2025, 8:30 – 12:30

Science learning is a highly relevant topic for the investigation of cognitive development. First, misconceptions about scientific phenomena follow a developmental trajectory. Second, science learning depends on metacognitive skills such as inhibition or cognitive reflection, that develop during childhood. Therefore, methods for investigating science learning can provide novel insights into cognitive development. The present tutorial introduces such a method, namely Bayesian modeling of science learning.

Bayesian inference is a method of statistical inference, i.e., a method of drawing conclusions from data. The central notion of this method is Bayes' theorem, which states that the probability that a hypothesis is true given some evidence (P(H|E), called posterior) is proportional to the product of the probability of the hypothesis (P(H), called prior) and the probability that the evidence would occur if the hypothesis were true (P(E|H), called likelihood). While Bayesian inference is a common method in empirical research, it can also be seen as a formalized model of science learning. The learner's prior belief (say, that heavier objects displace more water) is updated after seeing some evidence (say, that a lighter but larger object displaces more water). This approach is supported by previous research (Ullman & Tenenbaum, 2020; Colantonio et al., 2022).

The newly developed Python package "Bayesian Science Learning" (BaSciL) provides an easy-to-use and flexible tool for Bayesian modeling of science learning. It can be used in cases where children learn about scientific phenomena from experimental comparisons. Researchers upload a spreadsheet with information about experimental trials and participants' responses. The program automatically computes likelihoods of observed outcomes and learners' prior belief distributions and performs Bayesian updating. Because the functions are pre-built and run automatically, BaSciL makes Bayesian modeling accessible even to researchers unfamiliar with the Python programming language.

Aims:
The aim is that participants…

  • understand why Bayesian inference provides a formal model of science learning
  • are able to create an input table for the modeling in Python
  • know how to execute the package functions to perform the modeling
  • can inspect and interpret the modeling output
  • know about common problems during modeling

Prerequisites:
Basic knowledge on Bayesian statistics and a Laptop with Python installed

Participant limit: 12

T02: An introduction to Linear Mixed-effects Models in R
Organizer: Cintia Bali, University of Pécs, Department of Cognitive and Evolutionary Psychology

Thursday, 9 January 2025, 8:30 – 12:30

Using Linear Mixed-effects Models (LMMs), also known as random effects models, to analyze data can be highly beneficial. However, their use is not widespread, partly due to the considerable variation in how LMMs are applied and results are reported, making it difficult to understand them better. In comparison to traditional methods like ANOVAs, using LMMs offers more flexibility as they can handle complex data structures, missing data, and unequal datasets. LMMs also allow for the inclusion of random effects, enabling control for potential confounding effects, such as variability between participants. These features can lead to a better understanding of the data and more accurate conclusions. Therefore, the tutorial aims to introduce LMMs and provide a framework for building models and reporting results in a standard manner.

Aims:
At the end of the tutorial, participants will comprehend when and how to perform Linear Mixed-effects Models in R, interpret and present the outcomes of LMMs, and create tables and figures of the results in R.

Prerequisites:
Participants will need their notebooks and R Studio installed on them. Before the tutorial, participants will have access to the materials for the hands-on practice (such as the database or the script to follow through). No prior knowledge of R or explicit experience in programming is required to follow the tutorial.

Participant Limit: 25

T03: Analyzing Pupillometric Data in R: A Hands-On Tutorial
Organizers: Francesco Poli, University of Cambridge, Tommaso Ghilardi, Birkbeck University of London

Thursday, 9 January 2025, 8:30 – 12:30

Pupillometry—the measurement of pupil size and reactivity—is a valuable method in cognitive science for inferring cognitive and affective processes in both humans and animals. Changes in pupil size can provide insights into attention, arousal, and cognitive load, making it a powerful tool in developmental and comparative cognition research. However, analyzing pupillometric data presents challenges due to its complexity, including issues like missing data from blinks, noise, and the need for specialized statistical techniques.

This tutorial aims to demystify the process of analyzing pupillometric data using R, focusing on the PupillometryR package—a comprehensive toolkit designed to facilitate the cleaning, processing, and statistical analysis of pupil data. Participants will learn how to handle common issues such as missing data and noise, perform smoothing and filtering, and apply advanced modeling techniques like Generalized Additive Models (GAMs) which are particularly suited for time-series data inherent in pupillometry.

By the end of the tutorial, participants will have a practical understanding of how to preprocess and analyze pupillometric data in R. The hands-on approach ensures that attendees can apply these techniques to their own research, enhancing the rigor and depth of their data analysis in developmental and comparative cognition studies.

Aims:

  1. Understand the fundamentals of pupillometry and its applications in cognitive research.
  2. Learn how to import and preprocess pupillometric data in R using the PupillometryR package.
  3. Gain skills in data cleaning, including handling missing data due to blinks and other artifacts.
  4. Apply smoothing and filtering techniques to prepare data for analysis.
  5. Explore statistical modeling methods suitable for pupillometric time-series data.
  6. Develop proficiency in visualizing and interpreting pupillometric data and analysis results.
  7. Acquire the confidence to apply these techniques to their own datasets in future research.

Prerequisites:

  • Laptops with R and RStudio installed. Instructions for installing the required R packages (including PupillometryR) will be provided ahead of time.
  • Basic familiarity with R is beneficial but not mandatory. All code and detailed instructions will be provided during the tutorial.
  • No prior experience with pupillometry data analysis is required.

Participant Limit: 25