News and Events
The GRAL Research Group organizes different seminars and workshops related to Applied Linguistics and Second Language Acquisition. Between 2013 and 2016, the GRAL Research Group organized the following events:
Other events organized by the GRAL Research Group include talks and seminars by renowned professors and researchers. Please, find a list of the last talks below:
- Roeland van Hout (CLS, Radboud University Nijmegen) - "Why linear mixed models in applied linguistics?" (21st September 2016)
Many statistical analyses seem to have become obsolete or old-fashioned ever since linear mixed models are gaining ground in many branches of empirical linguistics. Simultaneously, people in the field involved started saying that it is important to do real statistics and that we can do so because we have R. Remarkably, linguists were fairly late after all in picking up the possibilities of linear mixed modelling (late adopters) and the familiar statistical packages kept their position in other domains of scientific research, even in using linear mixed modelling. What is the state of affairs?
In this presentation I want to explain what the advantages and disadvantages are of linear mixed modelling, not by showing the application of impressive, dazzling models, but by discussing and showing the basic concepts that can be dealt with because of linear mixed models:
1. What are random intercepts? Are they new? Why is this concept so popular in corpus linguistics and psycholinguistics? Has a simple t-test random intercepts?
2. What are random effects and how and why to distinguish them from fixed effects?
3. What are random slopes? What is the advantage of including them in statistical testing? And why is there so much discussion about including slopes or not? I will explain random slopes on the basis of a paired t-test.
4. What is nesting and what is hierarchical modelling? Why is this concept important when schools and classes are involved, as often happens in applied linguistic research?
5. What is cross-classification of random effects and how useful is it in applied linguistic research? I want to discuss this concept on the basis of my own research on L1 and L2 effects on L3 learning.
6. Why is model selection so crucial in linear mixed modelling, even more than the significance of individual effects? Why does it mean to keep models maximal or minimal?
I aim to address (a large part of) these questions on the basis of simple examples. My aim is to give you a better understanding why using linear mixed models may be appropriate and profitable in your own research or not.
(to access the presentation, please click HERE)
