The data were reviewed for inconsistencies and out of
range responses, edited, and weighted, using the complex
survey analysis module of SPSS Version 18. Weights were
computed as a product of three components: base weight,
which was an inverse of the final probability of selection,
adjustment for nonresponse at both the household and individual level, and post-stratification adjustment based on
residence (urban or rural), age, and gender from the 2008
population projection for Thailand. A two sample test was
used for pair comparison of prevalence among different
group of users at statistically significant level of p < 0.05.
Analyses of knowledge, addiction, and quitting, as well as
multivariate analysis, were carried out for men only,
because the levels of use of the tobacco products among
females were too low for reliable analyses. Multivariate analyses, taking the complex sample design into account, used
logistic regression models to see if the type of cigarettes
smoked (manufactured cigarette only, RYO only and dual
use) was associated with age, education, income, residence,
and region. These variables were selected because of previous studies that have shown these variables to be related to
tobacco use and the use of RYO. In the multivariate
models, we determined the relationship of one variable
while controlling for the effects of the others. The test for
importance of a factor was carried out by a comparison of
the full model with all factors and a reduced model where
the factor of interest was excluded