Heckman-type selection models to obtain unbiased estimates with missing measures outcome: theoretical considerations and an application to missing birth weight data

Table 5 Associations between IFAS and low birth weight

Logit Models					Heckman Models
Variable	Unadjusted		Adjusted		Unadjusted		Adjusted
	Risk differential		Risk differential		Risk differential		Risk differential
	dy/dx	95% CI	dy/dx	95% CI	dy/dx	95% CI	dy/dx	95% CI
IFA: No⁽⁺⁾
Yes	0.076	−0.101; 0.253	0.061	−0.184; 0.305	0.003	−0.026; 0.031	− 0.104***	− 0.169; − 0.039
Educational attainment: No schooling⁽⁺⁾
Primary school			0.064	− 0.155; 0.282			0.004	− 0.027; 0.035
Coranic			0.016	−0.393; 0.425			−0.010	− 0.069; 0.049
Secondary or higher			− 0.239	− 0.574; 0.096			− 0.053**	−0.101; − 0.005
Maternal age (years): 20–34⁽⁺⁾
15–19			0.898***	0.655; 1.141			0.126***	0.085; 0.167
35–49			0.569***	0.263; 0.875			0.051**	0.004; 0.099
Socioeconomic status: Most poor⁽⁺⁾
Poor			−0.161	−0.488; 0.167			−0.033	−0.078; 0.013
Middle			−0.225	− 0.549; 0.100			− 0.076***	−0.126; − 0.025
Rich			0.004	− 0.313; 0.321			− 0.051*	−0.102; 0.000
Most rich			−0.178	− 0.561; 0.205			− 0.074**	−0.133; − 0.016
Child sex: male
Female			0.338***	0.147; 0.529			0.045***	0.019; 0.070
Previous births: 0 child ⁽⁺⁾
1–4			−0.473**	−0.862; − 0.085			−0.065**	− 0.125; − 0.005
≥5			−1.283***	−1.782; − 0.785			−0.148***	−0.223; − 0.072
Twin births: No⁽⁺⁾
Yes			2.483***	2.199; 2.767			0.336***	0.288; 0.385
Constant	−1.987***	−2.115; − 1.859	−2.287***	−2.873; − 1.700
LR test (Rho = 0)							14.59***
observations	4510		4510		4510		4510

*** p < 0.01. ** p < 0.05.* p < 0.1;⁽⁺⁾ Reference category; dy/dx = Marginal effect is a change in the probability of Y = 1 with a unit change in X. Adjusted model controls for village fixed effects; CI: confidence interval

ISSN: 1471-2288