Modeling Mood Variation and Covariation among Adolescent Smokers: Application of a Bivariate Location-Scale Mixed-Effects Model
journal contributionposted on 14.08.2016, 00:00 by Oksana Pugach, Donald Hedeker, Melanie J Richmond, Alexander Sokolovsky, Robin Mermelstein
INTRODUCTION: Ecological momentary assessments (EMAs) are useful in understanding both between- and within-subject dynamic changes in smoking and mood. Modeling 2 moods (positive and negative affect, PA and NA) simultaneously will better enable researchers to explore the association of the mood variables and what influences them at both the momentary and subject level. METHODS: The EMA component of a natural history study of adolescent smoking was analyzed using a bivariate location-scale mixed-effects model. The proposed model separately estimates the between- and within-subject variances and jointly models the 2 mood constructs. A total of 461 adolescents completed the baseline EMA wave resulting in 14,105 random prompts. Smoking level, represented by the number of smoking events on EMA, entered the model as 2 predictors: one that compared nonsmokers during the EMA week to 1-cigarette smokers and the second one that estimated the effect of smoking level on mood among smokers. RESULTS: Results suggest that nonsmokers had more consistent positive and negative moods compared to 1-cigarette smokers. Among those who smoked, both moods were more consistent at higher smoking levels. The effects of smoking level were greater for NA than for PA. The within-subject association between mood constructs was negative and strongest among 1-cigarette smokers; the within-subject association between positive and negative moods was negatively associated with smoking. CONCLUSIONS: Mood variation and association between mood constructs varied across smoking levels. The most infrequent smokers were characterized with more inconsistent moods, whereas mood was more consistent for subjects with higher smoking levels.