Table 3.
Random Intercept Logistic Regression of Complete Home Smoking Bans on Predictors in San Diego, Tijuana, and Guadalajaraa (n = 1,554)
| Predictors | Model 1 | Model 2 | Model 3 | ||||||
|---|---|---|---|---|---|---|---|---|---|
| OR | 95% CI | OR | 95% CI | OR | 95% CI | ||||
| Tijuana resident | .138 | .088 | .216 | .139 | .088 | .218 | .229 | .134 | .389 |
| Guadalajara resident | .048 | .030 | .074 | .046 | .029 | .073 | .119 | .067 | .209 |
| Age | .992 | .983 | 1.001 | .997 | .987 | 1.007 | .990 | .979 | 1.002 |
| Gender | .735 | .548 | .986 | .778 | .579 | 1.047 | .916 | .652 | 1.285 |
| Employed | .867 | .721 | .995 | .862 | .733 | 1.014 | .947 | .775 | 1.002 |
| Education | .818 | .758 | .883 | .823 | .763 | .889 | .753 | .686 | 1.285 |
| Single marital status | .592 | .450 | .780 | .691 | .514 | .927 | 1.033 | .738 | 1.157 |
| Children | 1.625 | 1.179 | 2.241 | 1.920 | 1.349 | 2.732 | |||
| Nonsmoker | 2.271 | 1.499 | 3.441 | ||||||
| Former smoker | 1.574 | .947 | 2.616 | ||||||
| Other smoker in home | .503 | .364 | .696 | ||||||
| Acculturation to United States | 1.538 | 1.177 | 2.011 | ||||||
| SHSe aversion | 2.636 | 1.637 | 4.243 | ||||||
| Antitobacco society pressure | 1.123 | 1.088 | 1.159 | ||||||
| Work bans | 1.226 | .831 | 1.808 | ||||||
| Constant | 92.399 | 39.673 | 215.201 | 46.755 | 18.137 | 120.529 | .133 | .024 | .729 |
| χ2 (7) = 192.81, p < .001 | χ2 (8) = 197.62, p < .001 | χ2 (15) = 245.31, p < .001 | |||||||
Notes. CI = confidence interval; OR = odds ratio; SHSe = secondhand smoke exposure. The difference in fit from Model 1 to Model 2 is statistically significant χ2 (1) =197.62 − 92.81 = 4.81, p < .05. The difference in fit from Model 1 to Model 3 is statistically significant χ2 (8) = 245.31 − 192.81 = 52.5, p < .001. The difference in fit from Model 2 to Model 3 is statistically significant χ2 (7) = 245.31 − 197.62 = 47.69, p < .001. Bold-faced values are statistically significant across all models in which they were included.
Numbers in cells are ORs adjusted for other covariates, associated two-sided probabilities, and 95% CI based on random intercept logistic regressions of complete home ban. Total sample N was 1,103 in San Diego, 398 in Tijuana, and 400 in Guadalajara, although the N for analysis was 1,554 after missing data were deleted in 374 sample clusters. Analysis was conducted using mixed model employing random intercepts based on primary sampling units. Rho, the intraclass correlation within each analysis, was equal to .04. A random intercept approach was used in analysis using primary sampling units. Cities of residence were dummy coded with San Diego serving as referent in this model.
aModel 1 includes demographics. Model 2 adds presence of children to Model 1. Model 3 adds the additional potential mediators to Model 2.