Terminology issues
SLiM as a starting point
- ignore cluster
- model every cluster
- Why these aren't the best option
Visual starting point
- spaghetti plot
Random intercepts
- issues in interpretation
- prediction (conditional vs. not)
Random slopes
- additional interpretation
- compare to results from modeling every cluster
- prediction visualization
Simulation demo
Other Complexities
- Additional clustering, nested or otherwise
- Residual correlation
- glmm, stars
- Bayesian
Alternative approaches
- fixed effects, gee, sem
Exercises