|
Abstract:
|
Oakes raises many formidable issues about causal inference that remain challenging for public health researchers, although many are not unique to estimation of neighborhood effects. No overall solution to identification problems is realistic in this or any setting, but some incremental improvements can be made in the conduct and interpretation of observational multilevel research. Because a common focus is on the decomposition of effects into compositional and contextual components, clarification of this strategy is a worthwhile goal. Due to problems discussed by Oakes, such as omitted variables and emergent effects, it may be better to conceptualize the estimated contextual effect as a kind of residual effect, not the effects of measured predictors at the individual level.
|
