Socioeconomic inequalities in self-perceived oral health among adults in Chile

Table 3 Self-perceived oral health regression results

	OLS		Heckman Two-step		Ordered Model		Ordered Model Sample Selection
	Coef.	p-value	Coef.	p-value	Coef.	p-value	Coef.	p-value
Sex	−3.991	0.000	−2.433	0.042	−0.351	0.000	−0.092	0.411
Age (base: 18–30)
31–40	3.912	0.001	2.807	0.029	0.338	0.000	0.144	0.197
41–50	5.970	0.000	4.622	0.001	0.515	0.000	0.265	0.032
51–61	7.219	0.000	6.429	0.000	0.637	0.000	0.447	0.000
Head of household	1.522	0.038	−1.506	0.456	0.161	0.008	−0.270	0.049
Married/partner	−0.364	0.609	−0.186	0.784	−0.030	0.636	−0.002	0.973
Children under 5	2.273	0.001	2.974	0.000	0.196	0.001	0.265	0.000
Children aged 5-18	0.253	0.327	0.498	0.161	0.016	0.497	0.047	0.053
Education (base: Middle or less)
Incomplete secondary school	−1.184	0.217	−2.162	0.069	−0.113	0.182	−0.231	0.004
Complete secondary school	−3.455	0.001	−4.749	0.000	−0.320	0.001	−0.450	0.000
Higher Education	−6.031	0.000	−7.206	0.000	−0.536	0.000	−0.623	0.000
Employed full time	0.211	0.739	1.420	0.196	0.004	0.941	0.170	0.018
Employed part time	−1.018	0.251	−1.425	0.124	−0.116	0.130	−0.154	0.048
Healthcare system (base: Public Insurance A, most vulnerable)
Public Insurance B	−2.493	0.002	−3.166	0.001	−0.231	0.000	−0.291	0.000
Public Insurance C	−2.164	0.039	−3.640	0.004	−0.209	0.021	−0.379	0.000
Public Insurance D (least vulnerable)	−2.382	0.006	−3.475	0.007	−0.223	0.003	−0.341	0.000
Public Insurance unknown category	−2.131	0.172	3.431	0.771	−0.262	0.053	0.518	0.204
Inverse Mills ratio (IMR)			−4.294	0.105
Constant	32.623	0.000	42.148	0.000
Observations	1,394		1,394		1,394		1,394
R-squared	0.138
LR test indep equations							6.04	0.014

Abbreviations: OLS Ordinary Least Squares, Coef estimated coefficient
The dependent variable is the OHIP-14 score. Independent variables are described in Table 1. The first two columns report OLS estimates without correction for sample selection. The next two correct for sample selection using Heckman’s two-step estimator. The third set estimates an ordered multinomial model, whereas the final set corrects the ordered multinomial model for sample selection. Clustered standard errors at the municipality of residence level are reported
Inv. Mills ratio (IMR) is the sample selection correction variable in the Heckman model. After estimating the participation equation (Additional file 1: Table S1), for each observation in the sample we compute \( \hat{\lambda} \) =ϕ(x₃ \( \widehat{\beta_3} \))/Φ(x₃ \( \widehat{\beta_3} \)), where ϕ is the density of the Normal Distribution and Φ is its cumulative
LR test indep equations (likelihood-ratio test) tests hypothesis that the errors of OHIP-14 and participation equations are uncorrelated in the ordered response model

ISSN: 1475-9276