Table 6.

Logistic regression predicting W3 substance use based on change in parental movie restrictions from W1 to W2 among those who are substance-naïve at W2

		Full drink (n = 615)			Marijuana use (n = 627)
W1 restriction level	Change from W1 to W2	Total N	% onset W2–W3	Adjusted odds ratioa [95% CI]	Total N	% onset W2–W3	Adjusted odds ratioa [95% CI]
Fully restricted	Group 1: no change	21	6.25%	0.18 [0.02, 1.50]	22	21.43%	1.26 [0.25, 6.36]
	Group 2: more lenient	33	20.00%	0.39 [0.10, 1.54]	36	0%	–
Partially restricted	Group 3: no change	91	29.73%	0.86 [0.44, 1.67]	101	14.29%	1.59 [0.68, 3.69]
	Group 4: more restrictive	4	75.00%	–	4	0%	–
	Group 5: more lenient	90	32.00%	1.05 [0.54, 2.04]	91	19.72%	1.42 [0.61, 3.32]
Unrestricted	Group 6: no change	234	31.72%	(Reference)	233	13.48%	(Reference)
	Group 7: more restrictive	17	21.43%	0.38 [0.08, 1.76]	18	20.00%	2.07 [0.47, 9.16]
Pairwise comparisons of change within baseline restriction level groups
			1 v 2: Wald (1) = 0.38, p = 0.54
			3 v 5: Wald (1) = 0.25, p = 0.62			3 v 5: Wald (1) = 0.05, p = 0.82

^aAll models control for age, sex, race/ethnicity, lunch subsidy, school cohort, parental knowledge (three subscales), exposure to substance use (alcohol model only), average screen time, availability of substance, peer use, sensation seeking, and delinquency. Dashes indicate that the test was inestimable due to low or zero cell size counts. Cohort 2 is removed from these analyses because of contemporaneous measurement of W2 and W3 data for that cohort. Of those participants who were alcohol-naïve at W2, 125 were missing either W1 or W2 parental restriction data. Of those participants who were marijuana-naïve at W2, 122 were missing either W1 or W2 parental restriction data