Multilevel Modeling

3. Multilevel Framework

<a href="/longdesc/Subu_Sec3_Exer1.html">Flash Description</a>

Flash is not available on mobile devices. Please view the Flash Description.

Answering the question in Exercise 1 requires a study-type {y, T}, i.e., in which an ecological exposure (e.g., proportion in poverty) is linked to an individual outcome (BMI). A more complete representation would be type {y, x, T}, whereby we have an individual outcome, individual confounders (x), and ecologic exposure reflecting a multilevel structure of individuals nested within ecologies. When the ecological exposure is an aggregate measure of individual characteristics, such as percent poverty, it is obvious that information on both individual poverty and neighborhood percent poverty is required to test for an ecological effect. However, multilevel data are essential even if the ecological variable is a structural feature such as neighborhood presence of fast food outlets, because people with individual level disadvantage are likely to be overrepresented in places with structural risk factors.