Analysis of longitudinal ordinal data using semi-parametric mixed model under missingness

In studies related to social or medical sciences, ordinal responses are often recorded repeatedly over time on a subject. A semi-parametric model with spline smoothing has been considered to capture the temporal trend exhibited in the longitudinal data. In addition, information on covariates and/or responses may not be available in one or more visit. A dynamic model for both missing responses and covariates is considered here. The parameters are estimated by adopting MCNREM methodology. A detailed simulation study has been performed to justify the utility of the proposed model. The model is applied on the Alzheimer's Disease Neuroimaging Initiative (ADNI) data.