I'm planning to conduct a secondary data analysis of a large dataset collected from 10th grade students in 32 schools over
a 6-year period from 1999-2005. I'm interested in using multilevel models to predict the students' body mass index and blood pressure from
individual-level demographic and health behavior data, as well as school-level SES and indicators of a stressful environment.
While data were collected from a core set of 32 schools in each of the 6 years, the students participating in each school each year were unique, as the 10th grade was always the group assessed. Thus, I do not have repeated measures for the students at Level 1, but rather the schools were repeated. It is probable that both the school attended and, to a lesser extent, the year in which the students were assessed could impact the outcome. I'm unsure on how to specify my level 2 unit. Should it be the school? If so, how do I account for the repeated school information over time? Should it be the school/year combination? If so, how do I account for the fact that some of the schools were the same? In general, how can I account for the effect of time when repeated measures occur at Level 2, but not Level 1?
I look forward to your responses.