Table 3. Comparison of unexplained deviance in disparities in multi-variable models (N = 485 counties).

		Model
Model Fit		M0: Unadjusted	M1: Adjusted for Region & Urbanicity Effect Modifiers	M2: M1 + Homogeneous (Non-Race/Ethnicity-Specific) Factors1	M3: M2 + Race/Ethnicity-Specific Factors2	M4: M2 + Non-Race/Ethnicity-Specific Effect Modifiers1	M5: M3 + Race/Ethnicity-Specific Effect Modifiers2
df Error		965	948	928	927	908	906
BIC		10825	10385	10198	10195	10173	10092
Mean3		0.86	0.86	0.86	0.86	0.86	0.87
Adjusted PRR		2.36	2.36	2.35	2.36	2.37	2.4
Residual Variance		0.41	0.17	0.18	0.18	0.11	0.09
Total Unexplained Deviance4		1.15	0.91	0.91	0.91	0.86	0.85
% Unexplained Deviance 5		64%	81%	80%	81%	87%	90%
Unexplained Deviance Across Models (Diagonals) and Percentage of Deviance Explained by Effect Modifiers (Off-Diagonal)	M0	1.15	21%	21%	20%	25%	26%
	M1		0.91	0%	-1%	5%	6%
	M2			0.91	0%	6%	6%
	M3				0.91	6%	7%
	M4					0.86	1%
	M5						0.85

^aModels labeled with factors include just their main effects; models labeled with effect modifiers include main effects and interaction terms with ethnicity.

¹ Includes the following factors: Female prevalence rate, % IDU, New diagnosis rate, GINI index, % high school education, income inequality ratio, social association rate, violent crime rate, % severe housing problems, % rural, drug overdose mortality rate, Gonorrhea rate, % uninsured, preventable hospitalization rate, healthcare costs, Latino population % change, % Puerto Rican, % not English proficient; % Latino living below poverty.

² Includes the following factors: Population Ratio (% Latino/ % NL-White).

³ "Mean" coefficient represents the difference between Latino and NL- White log (PRR) after adjusting for county-level factors.

⁴ “Total Unexplained Deviance” is the sum of “Mean” and “Residual Variance”.

⁵ Percent unexplained deviance due to difference between Latino and NL- White log (PRR).