10. BCN Advanced (non)linear regression techniques in R
PhD students. Note that this course may also be taken instead of the compulsory BCN Statistics Course. However, in that case it is required that the student has an excellent grasp of how to conduct and interpret the results of a multiple regression analysis.
Many PhD students have to analyze complex datasets which involve repeated measurements, which are sometimes collected over time. In this course you will learn how to take into account this structural variability in your data using mixed-effects regression (also known as hierarchical regression or multilevel modeling). In contrast to repeated-measures ANOVA, the advantage of this approach is that no (perfectly) balanced dataset is required. Furthermore, as a regression approach, covariates can be easily included. Besides focusing on linear relationships, we will also focus on non-linear relationships between (combinations of) predictors and the dependent variable. To model these non-linear patterns, we will use generalized additive modeling. This approach in particular is very useful for analyzing any type of time series analysis (such as reaction time data, EEG data, eye tracking data, etc.).
Given that the teacher has a background in linguistics, the examples used in this course come from that field. However, the methods are easily applicable to data from other fields, and students are encouraged to bring their own data to work on during the lab sessions.
Between March 25 and April 4, 2019
This course consists of 5 lectures (10:00 - 14:00 hrs):
1. Linear mixed-effects regression using reaction time data
2. Generalized linear mixed-effects regression using eye tracking data
3. Generalized additive modeling (1D) using articulatory data
4. Generalized additive modeling (2D) using EEG data
5. Generalized additive modeling (nD) using geolinguistic data
Each lecture consists of about 2 hours of lecture followed by an additional 2 hours of lab session (students have to bring their own laptop). Lectures 1 and 3 are also included in the
BCN Statistics Course taught since 2016, so if students have followed this course, attending these lectures is not required (though still recommended as a refresher). The course will normally be taught on several days during a time span of two weeks.
Being able to use R, and knowledge about basic statistics (most importantly being able to conduct linear regression).
Martijn Wieling ( firstname.lastname@example.org )
|Last modified:||28 January 2019 3.27 p.m.|