Table 2.

Summary of bivariable, controlled, and mediated models: regression of consequences on depressive symptoms

Model	Men		Women
	Est.	P	Est	P
Bivariable	0.12	.09	0.13	.01
Controlled	0.09	.13	0.04	.35
Mediated
Total	0.14	<.001	0.05	.22
Direct	0.15	<.001	0.02	.66
Indirect	–0.01	.14	0.03	.13

Parameter estimates are regression (path) estimates from a multiple (sex) group model. All variables have been standardized to the between groups mean and standard deviation or centered at the between groups mean if dichotomous (race/ethnicity). P values refer to the test that the parameter is significantly non-zero. The bivariable model refers to a multiple (sex) group regression of alcohol use consequences on depressive symptoms. The controlled model refers to the regression of consequences on depressive symptoms adjusting for all other T1 variables (see Table 1). The mediated model is reported in full in Table 3 and refers to the regression of consequences on depressive symptoms, controlling for other T1 variables, and allowing for possible mediation via other T2 variables reflecting the freshman year. This full model has been trimmed of parameters that were not significant (within groups) at the α = .10 level in a backwards stepwise fashion