Table 10.

Standardized coefficient estimates from the spatial scale (SS) LARS/lasso model of BMI z-scores with no random effect (RE) in comparison to standardized coefficient estimates from three mixed models with a RE at the census block group (CBG), the census tract (CT), and the CBG and CT. The mixed models were fit using the covariates that were chosen by the SS LARS/lasso algorithm. The horizontal dashed line separates the individual-level variables and the neighborhood-level variables considered at multiple spatial scales.

Explanatory Variable	No RE	RE at CBG	RE at CT	RE at CBG and CT
Visit Age	0.077 (*)	0.081 (*)	0.081 (*)	0.081 (*)
Black	0.039 (*)	0.042 (*)	0.042 (*)	0.042 (*)
Distance to Medical Center	0.013 (+)	0.014 (+)	0.014 (+)	0.014 (+)
Population Density_CT	−0.034 (*)	−0.028 (*)	−0.028 (*)	−0.027 (*)
% Hispanic White_CBG	0.029 (*)	0.031 (*)	0.031 (*)	0.031 (*)
% Hispanic Black_CT	0.020 (*)	0.020 (*)	0.022 (*)	0.020 (*)
Median Household Income_CT	−0.052 (*)	−0.055 (*)	−0.056 (*)	−0.055 (*)
% Renter_CT	−0.023 (*)	−0.031 (*)	−0.030 (*)	−0.030 (*)
% Vacant_CT	0.013 (+)	0.015 (+)	0.016 (*)	0.015 (+)
Personal Crime Index_CBG	−0.010	−0.013	−0.016	−0.014
Property Crime Index_CT	−0.010	−0.012	−0.011	−0.012
Park Density_CT	0.008 (+)	0.010	0.010	0.010
Exercise Equipment_CBG	−0.035 (*)	−0.041 (*)	−0.038 (*)	−0.041 (*)

Notes: Values marked with (*) have a p-value < 0.05, and values marked with (+) have an associated p-value < 0.1. For the SS LARS/lasso model with no RE, covariates selected by the SS LARS/lasso algorithm were plugged into an OLS regression model to obtain estimated p-values.