A Census Data Based Simulation Experiment on the Effects of Modeling Neighborhood Effects on a Health Outcome at “the Wrong” Scale of the Census Geography

Claudia L. Nau, Pennsylvania State University

This paper presents a simulation experiment based on census data from Los Angeles County. The objective is to assess how predictions of neighborhood effects on a health outcome change when “neighborhood” is modeled at one scale of the census geography while the actual neighborhood effect is operating at another. Neighborhood effects (the effects of particular neighborhood characteristics on a health outcome) are simulated at three levels: block, block group, and tract. Within each level a set of scenarios is created by letting the strength of the neighborhood effect vary. Hierarchical linear models are used to assess (1) how the predictions of the neighborhood effect differ from the data generating parameter when the scale of the neighborhood used in the model is a) smaller or b) larger than that of the data generating process (2) how the effect of misrepresenting the neighborhood scale varies by strength of the underlying neighborhood effect.

  See extended abstract

Presented in Session 53: Peers, Neighborhoods, and Health