Multilevel Models, Equations, Syntax & English

Author
Affiliation

University of Michigan

Published

June 13, 2024

model equation Stata English
Intercept Only \(y = \beta_0 + e_{ij}\) mixed y We estimated the mean of [outcome]
Intercept
Independent Variable(s)
\(y = \beta_0 + \beta_1 x + e_{ij}\)
\(y = \beta_0 + \beta_1 x + \beta_2 z + e_{ij}\)
mixed y x
mixed y x z
We estimated the relationship of [independent variable(s)] with [outcome]
Intercept
Random variation of the intercept
\(y = \beta_0 + e_{ij} + u_{0j}\) mixed y || groupid: We estimated the mean of [outcome]. We allowed the intercept of the model to vary by [groupid].
Unconditional intraclass correlation coefficient (ICC) \(\frac{var(u_{0j})}{var(u_{0j}) + var(e_{ij})}\) mixed y || groupid:
estat icc
XX% of the variation in [outcome] was explained by clustering of participants in [groupid]
Intercept
Independent variable(s)
Random variation of the intercept
\(y = \beta_0 + \beta_1x + e_{ij} + u_{0j}\)
\(y = \beta_0 + \beta_1 x + \beta_2 z + e_{ij} + u_{0j}\)
mixed y x || groupid:
mixed y x z || groupid:
We estimated the relationship of [independent variable(s)] with [outcome]. We allowed the intercept of the model to vary by group.
Intercept
Independent variable
Random intercept
Random slope
\(y = \beta_0 + \beta_1 x + e_{ij} + u_{0j} + u_{1j} x\) mixed y x || groupid: x We estimated the relationship of [independent variable] with [outcome]. We allowed the intercept of the model to vary by group. We also allowed the relationship of [independent variable] with [outcome] to vary by group.
We can estimate multilevel models with more than 1 random slope. \(y = \beta_0 + \beta_1 x + \beta_2 z +\)
\(e_{ij} + u_{0j} + u_{1j} x + u_{2j} z\)
mixed y x z || groupid: x z