#random slope

Multilevel Models for Nested Data: Pupils in Classes, Classes in Schools

Surveying 800 pupils in 20 classes does not give you 800 independent observations. Ignoring nesting makes standard errors falsely small. ICC, the design effect,