The GYTS is a school-based survey, using a two-stage cluster sample design that produces representative samples of students in grades associated with ages 13 to 15 years. At the first stage, the probability of schools being selected is proportional to the number of students enrolled in the specified grades. At the second sampling stage, classes within the selected schools are randomly selected. All students in selected classes attending school the day the survey is administered are eligible to participate. Student participation is voluntary and anonymous using self-administered data-collection procedures. The GYTS sample design produces representative, independent, cross-sectional estimates for each site. For cross-site comparisons, data in this paper are limited to students aged 13 to 15 years old.
A weighting factor is applied to each student record to adjust for non-response (by school, class, and student) and variation in the probability of selection at the school and class levels. A final adjustment sums the weights by grade and sex to the population of school children in the selected grades in each sample site. SUDAAN [5
], a software package for statistical analysis of correlated data, was used to compute standard errors of the estimates and produced 95% confidence intervals, which are shown as lower and upper bounds. Differences in proportions were considered statistically significant at p
< 0.05, assessed by non-overlapping confidence intervals.
The GYTS enquired about several important tobacco-use indicators, including: current cigarette smoking (based on a response of "1 or more days" to the question, "During the past 30 days (1 month), on how many days did you smoke cigarettes?"); current use of tobacco products other than cigarettes; 'susceptibility' (that is, absence of a firm decision not to smoke) or likely initiation of cigarette smoking in the next year among never smokers (based on a negative response to the question, "Have you ever tried or experimented with cigarette smoking, even one or two puffs?" as well as a response of anything but "definitely no" to the questions, "If one of your best friends offered you a cigarette, would you smoke it?" and "Do you think you will try smoking a cigarette in the next year?") [6
]; exposure to cigarette smoke in public places (based on a response of "1 or more days" to the question, "During the past 7 days, on how many days have people smoked in your presence, in places other than your home?"); one or more parents smoke cigarettes (based on a response of "both", "father only", or "mother only" to the question, "Do your parents smoke?"); one or more best friends smoke cigarettes (based on a response of "most" or "all" to the question, "Do most or all of your best friends smoke?"); in favor of banning cigarette smoking in public places (based on a positive response to the question, "Are you in favor of banning smoking in public places (such as in restaurants, in buses, streetcars, and trains, in schools, on playgrounds, in gyms and sports arenas, in discos?"); and exposure to pro-tobacco advertising and promotion, either direct or indirect (based on: a response of "a lot" or "a few" to the questions, "During the past 30 days (1 month), how many anti-smoking media messages (for example, television, radio, billboards, posters, newspapers, magazines, movies, drama) have you seen or heard?", "During the past 30 days (1 month), how many advertisements for cigarettes have you seen on billboards?", "During the past 30 days (1 month), how many advertisements for cigarettes have you seen at point of sale?", "During the past 30 days (1 month), how many advertisements or promotions for cigarettes have you seen in newspapers or magazines?"; a positive response to the questions, "Do you have something (t-shirt, pen backpack, etc) with a cigarette brand logo on it?" or "Has a cigarette company representative ever offered you a free cigarette?").
-Tests were used to determine differences between subpopulations [7
]. Differences between prevalence estimates were considered statistically significant if the t
-value was < 0.05. Differences between prevalence estimates were considered statistically significant if the t
-value is associated with gender and that gender most often acts as an effect modifier for smoking and related risk factors. All analyses conducted in this study were gender stratified [8
The findings in this report are subject to at least three limitations. First, because the sample surveyed was limited to youths attending school, they may not be representative of all 13 to 15 year olds in Thailand. Second, these data apply only to youths who were in school the day the survey was administered and completed the survey. Student response was quite high in Thailand (99%; with a sample size of around 14,000), suggesting bias due to absence or non-response is small. Third, data are based on self-reports of students, who may under- or over-report their use of tobacco. The extent of this bias can not be determined in the Thailand data; however, responses to tobacco questions on surveys similar to the GYTS have shown good test-retest reliability [9