I have data of about 2000 children, with measures taken across 4 timepoints over the span of two years (2 measurement time points in first grade, 2 measurement time points in second grade). My baseline model is one in which time is nested within children nested within classroom. However, children are in different classrooms in first and second grade; so I am told I need to have a cross-classified model to account for the switching across classrooms. Is that possible to estimate? Is it done as a two- or three-level model?