Topics Covered
Nested Data
Understand why trials within subjects violate independence and inflate Type I errors.
Fixed vs. Random
Fixed effects estimate population means; random effects model individual variation.
Random Intercepts
Let each subject have their own baseline — the simplest mixed model and a huge step forward.
Key Concepts
- Why ignoring nesting structure leads to pseudo-replication
- The shrinkage effect: partial pooling vs. no pooling vs. complete pooling
- ICC (intra-class correlation) — how much variance lives between subjects
- When you need mixed models vs. when aggregation is okay
Homework
- Compute the ICC for a repeated-measures dataset
- Fit a random-intercepts model using lmer() and compare with a standard lm()
- Plot subject-level intercepts to visualize individual differences