This course will provide a practical guide and theoretical background of linear mixed effects modeling (LMER).
LMER is a statistical technique that has become more and more popular during the last couple of years. During the course, aspects related to exploratory versus confirmatory research, model selection, fixed and random effects, anova versus AIC/BIC, the optimization of statistical models, and model prediction will be discussed. The computational part of the course is based on numerous realistic data examples.
We will use R (R stdudio) as platform, but this course is not a course on R itself. Familiarity with R is assumed. |
|
The course will consist of two (intertwined) parts.
In the theoretical part we will discuss the ins-and-outs of LMER models, including the concept of significance, the "p-value", the way how these models try to optimize model parameters so as to minimize some error criterion, the balance between fixed and random terms, contrast coding, and the significance of predictors as a function of the presence of other predictors. We will also discuss blame analysis: properties of LMER models that, e.g., may help to understand which participants can be blamed for a certain effect, and touch on more advanced issues such as quantile regressions, and GAMS.
The practical part will be based on the application of various models on existing datasets or your own dataset. The course will start with simple examples. The platform we will use is R. |
|
Knowledge of R and of statistics 1 and 2 is assumed.
|
|
Knowledge of R and of statistics 1 and 2 is assumed.
|
|
The assessment takes place via a thesis.
|
|
|