Extended Data Fig. 5. Percentage changes in current smoking prevalence based on fixed effect coefficients from bivariate mixed effect linear regression models, by policy component, sex and age group.

Bivariate models examined the unadjusted association between smoke-free (P), health warnings (W), and advertising (E) achievement scores, and cigarette’s affordability (RIP) and current smoking prevalence, from 2009 to 2017, across 175 countries (n = 823 country-years). Linear mixed models were fit by maximum likelihood and t-tests used Satterthwaite approximations to degrees of freedom. P values were considered statistically significant if lower than 0.05.