Multilevel modelling
Hierarchical linear model with an ICC for nested scores.
When to use it
Use multilevel modelling when a numeric outcome sits inside groups and you want the intraclass correlation as well as the slopes. The ICC is the share of score variance that sits between groups. A high ICC means people in the same clinic resemble each other.
This procedure fits the same REML mixed model as the linear mixed model. The request uses different field names, and the report adds the ICC and a variance split.
Assumptions
The outcome and the level-1 predictors are numeric. The level-2 column has at least two groups. This fit uses REML and a random intercept. It does not test residual normality.
Running it in Tensr
Options
Prop
Type
Reading the output
The same linear mixed model, reported with the intraclass correlation and the split of variance between clinic and residual.
| Term | Estimate | Std. Err | p-value |
|---|---|---|---|
| const | 45.7742 | 2.35335 | p < .001 |
| hours | 0.973909 | 0.393446 | p = .013 |
| Level | Variance | % of total |
|---|---|---|
| Between clinic | 18.6857 | 47.07% |
| Within (residual) | 21.0128 | 52.93% |
Multilevel model for score, ICC = 0.471. Multilevel model for score, ICC = 0.471. Metrics: ICC = 0.471; Observations = 120; Groups = 12.
Reporting (APA 7)
Multilevel model for score, ICC = 0.471. Report the estimate in the table. This procedure is not summarised by one p-value.
Coming from SPSS
The product menu is Multivariate → Mixed Models → Multilevel Modelling (HLM). SPSS fits this with MIXED. The ICC is the ratio of the random-intercept variance to the total variance.
Related
The coefficient table without the ICC is the linear mixed model. A null-model ICC, z, and a confidence interval are on Mixed Model.