Modeling Smoking and Heaping Patterns in Self-Reported Cigarette Numbers by a Finite Mixture Approach

• Activity Number: 499
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Henry Yeh*+ and Byron Gajewski and Won S. Choi and Christine M. Daley
• Companies: University of Kansas Medical Center and Univ of Kansas-Medical Center and Univ of Kansas Medical Center, Dept. of Preventive Medicine and Public Health and Univ of Kansas Medical Center, Dept. of Family Medicine
• Keywords: finite mixture ; heaping ; inflation models ; negative binomial

Abstract

In smoking cessation research, heaping of self-reported cigarette numbers is a common problem because participants tend to report some rounded-off integers such as the multiples of 5, 10, or 20 instead of the exact counts. These excessive values show additional peaks in histograms and distort the conventional probabilistic distributions, e.g. Poisson or negative binomial, for count data analysis. Empirical evidence also suggests that in addition to heaping, particular patterns of smoking exactly half or one pack per day exist among smokers. In this work, we propose a finite mixture model for the actual latent count of cigarettes per day (CPD) to account for inflation of 10- and 20-CPD due to the personal smoking patterns, and a proportional odds model to connect the latent to the observed CPD for heaping behavior. Covariates e.g. age, gender, racial/ethnic groups are considered. We demonstrate this approach by analyzing data from the Center for American Indian Community Health (CAICH) to investigate American Indians' smoking behavior.