Hi, My model is testing neighborhood level effects on child outcomes, so children are nested within neighborhoods. The first research question looks at the child's home neighborhood, and the second research question looks at the child's school neighborhood and controls for students' home neighborhood characteristics.
I tried creating a unique cluster variable for child-by-school neighborhoods (so children residing in the same neighborhood and attending the same school are clustered together). However, using this cluster gives me an error that there is variation within cluster on my school variables.
Given that there is a higher ICC for the school cluster than the childxschool cluster, I'm inclined to cluster by school. What recommendations do you have?