Lecture 12: Multilevel Modelling (Alt 3) Flashcards
(27 cards)
What is Multi-Level Modelling (MLM)?
Multi-Level Modelling (MLM) is a statistical technique designed for hierarchically structured or nested data.
In which types of data is MLM commonly used in psychology and social sciences?
Data where individuals (e.g., students, patients) are nested within higher level units (e.g., classrooms, clinics, schools).
Which assumption of traditional regression models is violated by nested data?
The independence of observations assumption.
What is an example of a misleading conclusion from ignoring the nested structure in data?
Simple regression might falsely suggest that more therapy leads to worse outcomes.
What statistical phenomenon is illustrated when within-group and between-group patterns diverge?
Simpson’s paradox.
What are alternative names for Multi-Level Models?
Hierarchical Linear Models, Linear Mixed Models, Mixed Effects Models.
What does MLM do with variance?
Partition variance into components at different levels of the data structure.
What does allowing intercepts and slopes to vary by group mean in MLM?
They are treated as random effects.
What are random intercepts?
Different baselines across groups.
What are random slopes?
Effects of predictors can differ across groups.
How do MLM equations extend standard regression?
By including extra error terms that capture variation across groups and allowing both intercepts and slopes to be group-specific.
What do random intercepts in MLM account for?
Differences in baseline outcomes between groups.
What do random slopes in MLM allow?
The effect of predictors (e.g., therapy hours) to vary by group.
What are three benefits of MLM?
- Accurately modelling the data’s dependency structure
- Avoiding inflated Type I error rates from overstated sample sizes
- Preserving statistical power by avoiding aggregation or over-simplification.
What is lost when individual data is averaged to class or school level?
Variance is thrown away and important within-group patterns can be obscured.
What is the first stage in the general procedure for MLM?
Null model: A baseline with no predictors, testing whether there’s meaningful variation between groups.
What is added at Stage 2 of MLM modelling?
Level-1 predictors (e.g., student-level variables).
What is added at Stage 3 of MLM modelling?
Level-2 predictors (e.g., teacher or school-level variables) and test cross-level interactions.
How is model fit evaluated in MLM?
Using either model comparison methods (e.g., likelihood ratio tests, AIC) or the statistical significance of coefficients.
What makes MLM highly flexible?
It is suited to complex, hierarchical data structures, including multiple observations per subject, individuals nested within groups, and three or more levels of hierarchy.
In a repeated measures design, what are the typical Level 1 and Level 2 units?
Days or timepoints can be Level 1 and individuals Level 2.
Why is reporting MLM results more complex than standard regression?
Due to multiple levels of analysis and potentially many predictors across different levels.
What does the speed dating study illustrate in MLM?
How outcome variables at one level (perceptions of sexual interest) are predicted by variables at both the interaction and individual level (e.g., own sexual interest, sociosexual orientation).
Why is MLM essential in psychological research?
Because it is necessary when data points are clustered and non-independent.
