. Author manuscript; available in PMC: 2020 Jan 6.

Published in final edited form as: Child Abuse Negl. 2018 Dec 31;88:420–431. doi: 10.1016/j.chiabu.2018.12.022

Table 3.

Results from Sibling-Comparison Models

	Physical Health Problems^a	Depressive Symptoms^a	Educational Attainment^b	Income Attainment^b	Alcohol Problems^c	Antisocial Behavior^d

MIDUS ACEs
Between-Family Effect	.23	.45^*	−1.88^**	.45	.60	1.00^**
	(−.26 to .72)	(.07 to .83)	(−2.77 to −.99)	(−.31 to 1.21)	(−.31 to 1.31)	(.44 to 1.55)
Within-Family Effect	−.56^*	.51	−1.55	.96	.81	1.29^**
	(−1.09 to −.03)	(−.08 to 1.10)	(−3.99 to .46)	(−.64 to 2.55)	(−.29 to 1.92)	(.53 to 2.06)
Add Health ACEs
Between-Family Effect	.14^**	.30^*	−.90^**	−.73^**	.01	.63^**
	(.07 to .22)	(.05 to .55)	(−1.15 to −.65)	(−.97 to −.48)	(−.06 to .08)	(.42 to .83)
Within-Family Effect	.06	.53^*	−.29	−.31	.10	.16
	(−.04 to .16)	(.18 to .88)	(−.64 to .07)	(−.64 to .02)	(−.09 to .28)	(−.12 to .45)

Note: All models account for nested family structure of both samples.

Unstandardized regression coefficients presented with accompanying 95% confidence intervals presented in parentheses.

Models estimated using the MIDUS data include the following covariates: diet to control weight; smoked cigarettes in the past week; daily stressful experiences; verbal ability (BTACT); age; sex; and race. Models estimated using the Add Health data include the following covariates: physical activity in the past week; fast food consumption in the past week; smoked cigarettes in the past week; daily stressful experiences (perceived stress scale); verbal ability (Picture Vocabulary Test); age; sex; and race.

Missing values were handled using full information maximum likelihood (FIML) estimation for all estimated multivariate models.

Between-family effects provide an estimation of the association between adverse childhood experiences (ACEs) and each outcome across families. Within-family effects reflect differences in the same association between siblings from the same family.

Model estimated using a multilevel linear regression model.

Model estimated using a multilevel logistic regression model.

Model estimated using multilevel negative binomial regression model.

Model estimated using a multilevel linear regression model for the MIDUS sample and a multilevel negative binomial model for the Add Health sample.

p < .05;

^**

p < .001