Table 2.

Hierarchical Model Variances and Linear and Multiple Regressions. Fitted BLL model intercepts, slope coefficients and coefficient standard errors (SE), variance of household random effects ( ${\sigma }_{\mathit{house}}^2$), variance of residuals ( ${\sigma }_{\epsilon }^2$), coefficient of determination (R²), and mean absolute error (MAE) for (a) a hierarchical linear regression against total Pb in soil, (b) a hierarchical linear regression against bioaccessible Pb in soil, and (c) a hierarchical multiple linear regression against bioaccessible Pb and child age. In the hierarchical multiple linear regression, child age was not significant at the 95% confidence level.

Hierarchical model	Intercepta	Coef. (SE)a	${\sigma }_{\mathit{house}}^2$	${\sigma }_{\epsilon }^2$	R2	MAEb
(a) Total Pb	1.3	1.4 (0.42)	0.92	1.4	0.25	1.22
(b) Bioaccessible Pb	1.2	1.6 (0.44)	0.84	1.4	0.28	1.19
(c) Multiple regression	1.6		1.03	1.3	0.27	1.21
Bioaccessible Pb		1.5 (0.46)
Child age		−0.076 (0.13)c

^aUnits are 1000 mg Pb per kg soil.

^bMean Absolute Error (MAE); units are μg/dL.

^cChild age was not significant in the multiple linear regression model.