Estimated percentage change in smoking prevalence (the proportion of smoking in the general population) per 10% change in the exposure from ARIMAX models
| Smoking prevalence | Unadjusted | Adjusted for covariates in table |
| Percentage change per 10% change in the exposure (95% CI), p Value | Percentage change per 10% change in the exposure (95% CI), p Value | |
| Model 1 | ||
| Mass media expenditure (lag 0) | −0.03 (-0.09 to0.03), 0.275 | −0.03 (-0.09 to 0.02), 0.258 |
| Weekly spend tobacco (lag 1) | 0.08 (-0.09 to 0.25), 0.371 | |
| Cessation aid use (lag 1) | −0.18 (-0.38 to 0.03), 0.096 | |
| Tobacco control policies | −0.54 (-1.44 to 0.36), 0.238 | |
| Best fitting model | ARIMAX (0, 1, 1) (0, 0, 0) | ARIMAX (0, 1, 1) (0, 0, 0) |
| Non-seasonal (p)—AR | NA | NA |
| —MA | <0.001 | <0.001 |
| Seasonal (p)— AR | NA | NA |
| —MA | NA | NA |
| R2 | 0.465 | 0.527 |
| Model 2 | ||
| Mass media expenditure (lag 3) | −0.03 (-0.09 to 0.03), 0.269 | −0.03 (-0.09 to 0.03), 0.299 |
| Weekly spend tobacco (lag 1) | −0.08 (-0.10 to 0.25), 0.379 | |
| Cessation aid use (lag 1) | −0.18 (-0.39 to 0.03), 0.097 | |
| Tobacco control policies | −0.54 (-1.44 to 0.35), 0.235 | |
| Best fitting model | ARIMAX (0, 1, 1) (0, 0, 0) | ARIMAX (0, 1, 1) (0, 0, 0) |
| Non-seasonal (p)—AR | NA | NA |
| —MA | <0.001 | <0.001 |
| Seasonal (p)—AR | NA | NA |
| —MA | NA | NA |
| R2 | 0.465 | 0.527 |
The assumption of normally distributed errors was met. When lags of tobacco spending and cessation use were set to zero very similar results were found for mass media in model 1 (β=−0.03 (-0.09 to 0.02), p=0.238) and in model 2 (β=−0.03 (-0.10 to 0.03), p=0.278). Additional MA (0, 1, 2) or AR (1, 1, 1) terms were not significant.
AR, autoregressive terms; ARIMAX, Autoregressive integrated moving average modelling with exogenous variables;MA, moving average terms.