Zero-Inflated Ordinal Mixed-Effects Models

Excess zeros issues are very common in count data. Furthermore, when different type of zero sources are suspected in the data, zero-inflated models, such as zero-inflated Poisson model and zero-inflated Negative Binomial model, are commonly used in research. When the zero-inflation issue exists in the longitudinal data, the correlation between the observations nested within a cluster (subject) should be furthered considered. Zero-inflated mixed-effects Poisson model is one of the popular models to handle zero inflation and correlation within each cluster (subject) for count data. The excess zero issue can exist in the ordinal outcomes too. Kelley and Anderson (2008) proposed a zero-inflated-proportional odds model (ZIPO) for the ordinal outcomes with excess zeros, based on the framework from Lambert’s two-part model (Lambert 1992). In this thesis I proposed a Zero-Inflated Ordinal Mixed-Effects Model for the ordinal outcomes with excess zeros in longitudinal settings. The proposed model was tested using a series of simulations for model performance. The Zero-Inflated Ordinal Mixed-Effects Model was also applied to a real data analysis by Social Emotional Contexts of Adolescent and Young Adult Smoking Patterns (SECAP) data.