. 2020 May;32(2):765–778. doi: 10.1017/S0954579419000749

Table 2.

Results of the three-level meta-analysis models

	k studies	n	Effect Size		Test Statistic
	k studies	n	r	95% CI	Test Statistic
Random-effects model
Overall effect size	56	220	.15	[.12, .17]	t(219) = 12.94, p < .001
Mixed-effects models Substantive Moderators
Region					F(2, 208) = 1.01, p = .365
North America	35	138	.13	[.10, .16]
Europe	9	46	.16	[.12, .21]
Australasia	9	27	.17	[.12, .22]
Age of child					F(2, 215) = 0.62, p = .541
0 to 5 years	6	22	.18	[.11, .24]
6 to 11 years	8	25	.14	[.08, .20]
12 to 18 years	41	171	.14	[.11, .16]
Gender of child (% female)	41	181			t(179) = -0.16, p = .873
Gender of parent					F(2, 217) = 0.86, p = .426
Father	25	82	.16	[.13, .19]
Mother	16	59	.14	[.10, .18]
Both	31	79	.13	[.10, .17]
Parental substance abuse					F(3, 205) = 50.15, p = .002
Alcohol^a	21	83	.13	[.09, .16]
Tobacco^b	25	70	.13	[.10, .16]
Drug use^a,b,c,d	8	14	.25	[.19, .31]
Alcohol use disorder^c	11	42	.14	[.09, .19]
Child well-being dimension					F(2, 214) = 20.65, p = .073
Physical	38	162	.15	[.12, .17]
Psychological	16	42	.15	[.11, .19]
Cognitive	4	13	.06	[-.02, .14]
Methodological Moderators
Informant on parental substance abuse					t(212) = -10.08, p = .280
Parent self-report	47	159	.15	[.13, .18]
Child report	7	55	.12	[.06, .17]
Informant on child well-being					F(2, 203) = 10.53, p = .220
Child self-report	42	166	.13	[.11, .16]
Parent report	12	33	.17	[.12, .22]
Teacher report	4	7	.10	[.02, .19]
Shared informant					t(218) = 10.34, p = .182
Not shared	44	142	.14	[.11, .16]
Shared	17	78	.17	[.13, .21]
Mode parental substance abuse					t(169) = -0.29, p =.770
Interview	24	83	.15	[.12, .18]
Questionnaire	27	88	.14	[.11, .18]
Mode child well-being					t(174) = 0.02, p = .998
Interview	17	57	.14	[.11, .18]
Questionnaire	26	119	.14	[.11, .17]
Time lag association (months)	51	195			t(193) = -0.43, p = .667

Note. Some moderators were missing for certain studies. Each study can contribute multiple effect sizes, thus study sample size across subgroups can exceed the total study sample size for the Level 2 moderators. Within each moderator having more than two subgroups, identical letter superscripts indicate significant (p < .05) pairwise comparisons between subgroups.