“…Logistic regression models regressed smoking ban on each of the demographic and theoretical mediator variables described above to examine the association between the DV bans and mediator variables. Finally, the relative contribution of theoretical mediators to the likelihood of having a smoking ban in the home was estimated by random intercept regression models in order to adjust for cluster sampling and to improve fit due to variation in intercepts among sample clusters (Hovell & Hughes, 2009;Rabe-Hesketh & Skrondal, 2008). Models were examined to test whether theoretical mediators remained associated with the DV, after controlling for city and demographics, and to determine if reduction in the direct effect between the independent variable and the DV was observed when controlling for mediation variables, consistent with evidence of mediation (Baron & Kenny, 1986;Kraemer, Kiernan, Essex, & Kupfer, 2008).…”