Dissertations & Theses (Open Access)

Date of Award

Fall 12-2018

Degree Name

Doctor of Public Health (DrPH)


Sheng Luo, Phd

Second Advisor

Hulin Wu, Phd

Third Advisor

Momiao Xiong, Phd


In the slow progression of Parkinson's Disease (PD), impairments arise and affect multiple domains (e.g., motor, cognitive, and behavioral). Mixed types, multivariate longitudinal data are commonly used in PD studies. Challenges exist in assessing PD status and investigating disease progression due to lack of biomarkers and ubiquitous impairment in the disease. We proposed a model framework by combining the semi-parametric approach and multi- dimensional framework, and used the proposed model to investigate the heterogeneous disease development and the non-linear treatment effects in the multiple domains predefined in PD.

Furthermore, we extended the semi-parametric multidimensional approach to the data with multi-types endpoints. We investigated the multi-type events (competing risks) simultaneously with longitudinal profile in presence of impairment across domains and domain specific heterogeneous disease progression. Our approach provides an explicit framework for defining and estimating the impaired covariate effects, the association between domain specific longitudinal profile and multi-type endpoints.

Lastly, we addressed the missing data in PD. We extended the multidimensional joint model to missing data by analyzing two missingness patterns (intermittent and monotone missingness) jointly in domain levels. We provided a statistical method for simultaneous likelihood inference on missing data in presence of two missingness patterns and two missing mechanisms, missing at random (MAR) and missing not at random (MNAR). In summary, the studies in this dissertation add to current PD studies by focusing on those ignored or not fully addressed problems in PD. The applications in longitudinal data, survival data and missing data promote this framework usability in public health research.