Abstract
The process of transition from nonsmoking to regular weekly smoking during a period from 10 to 16 years was examined using data gathered during the course of a longitudinal study of 957 New Zealand adolescents. These data were analyzed using a latent Markov model to estimate both rates of transition between stages of smoking and errors of measurement in report data. The fitted model suggested that the process of transition to smoking was a progressive and one-way process in which once teenagers had graduated to a given smoking status, return to earlier stages was uncommon. This process also showed a clear tendency to accelerate with age, so that rates of transition to smoking behaviors after the age of 14 years were far higher than rates of transition at age 10 years. The model also made it possible to estimate the accuracy of smoking reports. This suggested that individuals who were nonsmokers or regular smokers were classified with better than 95% accuracy on the basis of report data. However, the reporting accuracy of occasional smoking was poor, with 42% of occasional smokers being falsely classified as nonsmokers. The implications of these findings for the understanding of the processes of transition to smoking behaviors and the effects of measurement errors in report data are discussed.