Missing data assumptions and methods in a smoking cessation study.