Modeling Direct Food Establishment and Latent Bivariate
Residential-Workplace Neighborhood Effects on Body Mass Index

James O’Malley, Ph.D.
Professor of Biostatistics
Department of Biomedical Data Science
The Dartmouth Institute for Health Policy and Clinical Practice
Geisel School of Medicine at Dartmouth

Tuesday, December 6, 2016
4:15 PM
Waterman 427A



Hierarchical modeling is the preferred approach of modeling
neighborhood effects. Given residential and workplace location
indicators, a bivariate (residential-workplace) neighborhood random
effect whose correlation quantifies the extent that a neighborhood's
residential effect correlates with its workplace effect may be
specified. However, statistical model estimation software typically does
not allow correlations between the effects of different clustering
variables. In this talk, I develop a Bayesian model with a bivariate
random effect for neighborhood and an accompanying estimation procedure.

The model accounts for individuals who reside or work in multiple
neighborhoods across their observations, individual heterogeneity
(random intercepts, random slopes), and serial correlation between
observations on the same individual. We apply the models to linked
Framingham heart study – food establishment data to examine whether (i)
Proximity to fast-food establishments is associated with Body Mass Index
(BMI); (ii) Workplace neighborhood exposure associations are larger than
those for residential exposure; (iii) Residential neighborhood exposure
associations correlate with workplace neighborhood exposure. For
robustness, we evaluate these hypotheses under multiple specifications
of the prior distribution of the neighborhood random effect covariance
matrix. In addition, we show that allowing for time varying neighborhood
membership, individual heterogeneity, and serial correlation across time
yields more precise neighborhood level estimates.

Math Department