Also known by hierarchical models, mixed models, among other names.
They are used to look at the effects of levels on the predictor variable.
Hierarchical, aka nested data (i.e. towns within counties, patients within hospitals, etc).
A good sample rate: although there is no rule of thumb for how many units within a level, the lower the number the higher the chance of errors in your model (i.e. 327 units within 12 divisions)
Level 1 random intercept model
Iij = δ00 + u0j + εij + α1j Xij.
Where Iij = δ00 + u0j + εij is the Null model
α1j Xij is the variation due to predictor x