9:00 – 9:10
Opening address
(Rebecca Borg & Pia Järnefelt)
Introductions, workshop goals and purpose. Reminders and announcements.
9:10 – 10:00
Getting to know our research
Research presentations (5 min presentations, 1-2 questions for each talk)
9.10 Erik Bojerud
9.30 Jinhee Kwon
9.40 Mattias Forsgren
10.00-10.30
Fika break
10.30-12:00
Getting to know our research
Research presentations (5 min presentations, 1-2 questions for each talk)
10.30 Yulia Kashevarova
10.40 Alireza Mahmoudi Kamelabad
10.50 Lilja Jónsdóttir
11.00 Emma Heeman
12:00 – 13:00
Lunch break
13:00 – 14:00
Strong tests of theories benefit from more informative analysis approaches:
From ANOVA to mixed-effects regressions
(Florian Jaeger)
Reading: Lohse, Kozlowski, & Strube (2023) -- at least the first 19 pages!
Mixed-effects or multi-level regression (aka Generalized linear mixed-effects models) increasingly is the default analysis approach across large parts of the linguistic and psychological sciences. This framework subsumes more traditional approaches, like t-tests or ANOVAs, is conceptually more transparent, and provides many additional advantages. I will go through an example data set to show how to prepare, conduct, and interpret linear & logistic mixed-effects regression.
If you've never used mixed-effects models before, you will probably find Bodo Winter's hands-on R-based introduction helpful (Tutorial 1 & 2). You should reserve 2.5h to work through these mostly math-free tutorials. It'll be worth your time, and you will gain a much deeper & practically-oriented understanding of generalized linear mixed-effects models. An R-based tutorial by Florian Jaeger covers additional details on predictor coding (centering, scaling of continuous factors; treatment vs. effects vs. Helmert vs. sliding difference coding).
We will only have one hour. So this class will only work if you are prepared. Please do your readings and submit questions about the reading(s) by 6pm the day before the session through the Slack channel #mixed-effects-models.
14:15 – 15:15
Methods session 1, Discussion: Can AI learn a language like a human baby?
How new data and methods are changing the way we build and test a theory about human language
(Chigusa Kurumada)
Syllabus here & Reading: Vong et al. (2024) Science.
Slides
15:30 – 17:00
Office hours (consult the experts)
(Florian Jaeger/Chigusa Kurumada)
9:00 – 10:00
Getting to know our research
Research presentations (5 min presentations + 1-2 questions for each talk)
9.10 Annika Høeg
9.20 Christoffer Forbes Schieche
9.30 Carla Wikse Barrow
10:00-10:30
Fika break
10:30-12:00
Strong tests of theories benefit from more informative analysis approaches:
From frequentist to Bayesian mixed-effect regression
(Florian Jaeger)
Reading: If you have familiarity with Bayesian approaches, consider reading Schad, Nicenboim, Betancourt, Bürkner, & Vasishth (2023). Otherwise Wagenmakers et al. (2018) provides an introduction to the general reasoning behind Bayesian approaches and its advantages over frequentist approaces. Both articles are long but worth the read. See how far you get. If either of these articles feels a bit overwhelming, consider reading an example cognitive science paper that uses Bayesian data analysis instead. Selfishly, I'll propose Tan & Jaeger (2024). This will give you an example of how one might report the results of a Bayesian mixed-effects regression.
During the meeting, we'll go through the same data as used on Monday to set-up, fit, and interpret a Bayesian mixed-effect logistic regression. The goal is to compare what changes, and what stays the same when moving from a frequentist to a Bayesian mixed-effect regression. Time permitting, I briefly sketch how the same Bayesian framework in R can be used to fit a much large range of models, including a variety of powerful models that should be of interest to linguists (e.g., multivariate regression, distributional regression for phonetic data analysis) and psychologists (e.g., psychometric models, mixture models for perceptual decision-making).
12:00 – 13:00
Lunch break
13:00 – 14:30
Methods session 2, Demo: A 4-part experimental design puzzle
We will focus on a participant's thesis study and learn about how to design an experiment
(Chigusa Kurumada)
Syllabus here
14:30 – 15:30
Office hours (consult the experts)
(Florian Jaeger/Chigusa Kurumada)
15:30 – 16:30
Guest lecture by Julie Sedivy: Melding the science of language with the art of writing
"Many of us may wish to meaningfully integrate our scientific knowledge with other important pursuits in our lives. This often requires cultivating an extreme form of interdisciplinarity, in which very different skill sets must be learned, and clashing values and approaches need to be resolved. I will offer a case study from my own efforts of writing a (forthcoming) book that involved weaving my experiences and insights as a psycholinguist into a literary memoir. I will outline some of the inevitable tensions that arose and ways in which I did my best to resolve them. I will then open the discussion to a broader exploration of how our experiences as language scientists might be integrated with other goals or priorities."
9:00 – 10:00
Getting to know our research
Research presentations (5 min presentations + 1-2 questions for each talk)
9.10 Maria Mystakidou
9.20 Amanda Kann
9.30 Rebecca Borg
9.40 Pia Järnefelt
10:00-10:30
Fika break
10:30-12:00
Methods session 3, Tutorial: "Research Life Cycle" & Pre-registration 101
We will discuss why and how to use an archiving system like OSF to make our studies "open"
(Chigusa Kurumada)
Syllabus here
12:00 – 13:00
Lunch break
13:00 – 15:00
Office hours (consult the experts)
(Florian Jaeger/Chigusa Kurumada)
15:00 – 16:00
Data visualization - Clarity and credibility
(Timo Roettger)
16:15 – 17:00
Office hours (consult the experts)
(Florian Jaeger/Chigusa Kurumada)
9:00 – 9:15
Student session: What happens next?
9:15 – 10:00
Symposium reflection: What did we learn?
10:00 – 10:30
Fika break
10:30 – 11:30
Continuation of Student session
11:30 – 12:00
Closing remarks