A latent class model of smoking experimentation in children

A latent class model to estimate the accuracy of reports of child smoking behaviour is presented. This model was fitted to data for a sample of 810 9-yr-old New Zealand children. The model suggested that errors of measurement in reports of child smoking largely arose from false negative responses in which children who had smoked described themselves as not smoking: false negative rates were in the region of 40%. The consequences of the high false negative rates are shown to lead to an underestimation of the prevalence of smoking experimentation and of the strength of association between maternal and child smoking.