Lecture 12: Multilevel Modelling Flashcards
(24 cards)
What does MLM stand for in statistics?
Multi-Level Modelling
What kind of data is MLM designed to analyse?
Hierarchically structured or nested data
What is a common assumption of traditional regression that MLM corrects?
Independence of observations
What paradox can arise from ignoring data hierarchy?
Simpson’s paradox
What can ignoring group-level clustering falsely suggest?
Misleading effects of predictors (e.g., therapy appears harmful)
What are random intercepts used for in MLM?
Adjusting for group-level baseline differences
What do random slopes allow in MLM?
Predictor effects to vary between groups
What does MLM do with variance in the data?
Partitions it across different levels
How does MLM help Type I error rates?
It avoids inflation by accounting for clustering
Why is MLM better than averaging group data?
It preserves within-group variance
What does MLM improve by modelling the data’s structure?
Statistical power and inference accuracy
What is the first model used in MLM analysis?
The null model
What does the null model test in MLM?
Whether there is meaningful between-group variance
What are level-1 predictors?
Individual-level variables
What are level-2 predictors?
Group-level variables
What can be tested by adding level-2 predictors?
Cross-level interactions
What methods can be used to evaluate MLM fit?
Likelihood ratio tests and AIC
What kind of data structure can MLM handle beyond two levels?
Multi-level hierarchies (e.g., students in classes in schools)
In repeated measures, what can be treated as level 1?
Time points or days
Why is reporting MLM more complex than standard regression?
It includes multiple levels and many predictors
What type of variables can MLM include in one model?
Both interaction-level and individual-level variables
When should MLM be used?
When data is clustered or non-independent
What major risk does standard regression face when ignoring hierarchy?
Producing misleading results
What does MLM provide for analysing hierarchical effects?
Richer and more accurate understanding
