Smoking on the shoulders of giants - An application of the two-part model to Swedish longitudinal data