. Author manuscript; available in PMC: 2017 Sep 1.

Published in final edited form as: J Res Adolesc. 2015 Mar 13;26(3):403–417. doi: 10.1111/jora.12200

Table 3. Multivariable Ordinary Least Squares Regression of Severity of Depression by Demographic Characteristics, Environmental Risks, and Substance Use.

	Model 1			Model 2			Model 3			Model 4

	b	SE	b*	b	SE	b*	b	SE	b*	b	SE	b*
Male	0.109	0.098	0.055	0.002	0.109	0.001	0.119	0.098	0.06	0.117	0.099	0.059
Age	0.055^*	0.023	0.129	0.064^**	0.024	0.149	0.057^*	0.023	0.132	0.054^*	0.023	0.126
Family maltreatment	0.260^*	0.108	0.145	0.250^*	0.112	0.14	0.304^**	0.108	0.169	0.253^*	0.109	0.141
Count of traumatic stressors	0.075^*	0.029	0.148				-0.036^a	0.057	-0.071	0.076^**	0.029	0.149
Categories of trauma
Direct physical violence				0.408^**	0.129	0.198
Direct sexual violence				-0.058	0.12	-0.026
Indirect exposure				-0.145	0.1	-0.075
Witnessing in person				0.072	0.116	0.04
Stressful life events	0.028	0.028	0.06	0.040	0.03	0.086	0.034	0.028	0.074	0.006	0.038	0.013
Substance use
Monosubstance use	0.320^*	0.126	0.143	0.271^*	0.127	0.121	0.361^**	0.123	0.161	0.316^*	0.126	0.141
Polysubstance use	0.438^**	0.146	0.174	0.379^**	0.145	0.149	0.425^**	0.146	0.169	0.444^**	0.146	0.176
Interaction term^b							0.151^*	0.065	0.24
Interaction term^c										0.048	0.055	0.069
R²	0.154			0.177			0.168			0.157
df	307			289			306			306
F	9.464			7.056			10.17			8.759

Note. b, unstandardized regression coefficient; b*, standardized regression coefficient; SE, standard error, calculated using robust estimator; df, degrees of freedom. Analyses were conducted using data from participants who had completed information for all variables included in the analytic model (n = 315); accordingly, the df reflects number of cases with complete data used in the analyses, less number of parameters estimated. The majority of missingness (80%) were due to missing data on traumatic stressors, family maltreatment, or a combination. Participants with missingness on these variables did not differ in age, gender, or severity of depression compared to participants with no missingness.

p < .05.

^**

p < .01.

^***

p < .001.

The interaction term between family maltreatment and traumatic stressor was highly correlated with the main effect of traumatic stressor, r(328) = .84, p < .001, resulting in a change of direction in the latter. According to Allison (2012), multicollinearity caused by the inclusion of an interaction term can be safely ignored.

Interaction term between family maltreatment and traumatic stressors associated with street life.

Interaction term between family maltreatment and stressful life events.