13 Multi-Level Models (or Hierarchical Linear Modelling, or Virtue in Danger)

Question 1

What is the point of hierarchical linear modelling?

Answer 1

If individuals’ scores are clustered, it allows you to analyse variance not only at the level of the individual, but also at the level of the clusters –e.g., the classroom.

Question 2

Hierarchically structured data is data in which _____-_____ units are nested within _____-_____ units.

Answer 2

Hierarchically structured data is data in which lower-level units are nested within higher-level units.

Question 3

What are three advantages of hierarchical linear modelling? (Assuming data is hierarchically structured)

Answer 3

1, Avoids confounding of effects at different levels by partitioning variability of DV into Level 1 and Level 2 components and modelling each level separately.

2. Explicitly addresses the dependence among lower-level observations that are nested within the same higher-level unit.
3. Provides a way of estimating separate regression coefficients for different Level 2 units and modeling cross-level interactions.

Question 4

In a multi-level model, what is always level 1?

Answer 4

The level at which data is collected.

Question 5

What would level 1 data be for a:

within-subjects design?

between-subjects design?

Answer 5

Within-subjects – all data points, clustered for each individual.

Between-subjects –raw scores for all children, say, clustered by classroom.

Question 6

Level 2 variables are ______ within the cluster, but may vary _____ clusters, causing systematic differences.

Answer 6

Level 2 variables are constant within the cluster, but may vary between clusters, causing systematic differences.

Question 7

In conventional regression, it is assumed that all variance comes from variation between individuals; in multi-level modelling it is assumed that….

Answer 7

Some variation comes from individuals and some comes from clustering.

Question 8

In Random Effects ANOVA, what are the random and fixed effects?

Answer 8

Random effects –error terms

Fixed effects –effect due to group

Question 9

What is an unconditional model?

Answer 9

A model in which the estimates are not conditional on any other (potentially mediating) variables.

Question 10

What does a Random Effects ANOVA do?

Answer 10

Tell us how much variance is due to within-subjects factors vs between-subjects factors.