Data were analysed using Stata SE 12.1 (StataCorp,
TX, USA). A series of 22 chi-square analyses
were conducted to examine whether ads of each
lifestyle topic were more or less likely than ads
about all other topics to: (i) feature particular message content, execution styles, emotional not physical
activitydepicted scene vs. not depicted scene);
and (ii) address key behavioural principles.
Poisson regression analysis was used to test for
linear changes in ad characteristics over time, accounting for the number of ads sampled each year.
Due to multiple testing, a conservative level of statistical significance (P