A SURVEY OF STATISTICAL METHODS FOR INVESTIGATING RISK OF LOW BACK PAIN IN A COHORT OF MANUFACTURING WORKERS
Longitudinal or panel studies involve taking repeated measurements on individuals and monitoring their status over time. Depending on the goals of the study, a variety of analysis methods may be appropriate for addressing specific research questions. This thesis presents a survey of statistical methods that can be used to analyze such data. Regression models for count data focus on the rate of occurrence of specific events. Survival analysis is particularly effective at modeling event relative risk in the presence of censored data. For repeated events of the same type, random subject effects can be included to take into account within subject variation. Multi-state models can characterize the entire event progression history when more than one status variable is tracked over time. We offer applications of each aforementioned method to a cohort study on the severity of low back pain in manufacturing and production industries. Assumptions and requirements of each method are discussed as well as inference and interpretation. This analysis is particularly useful for learning about these types of disorders as researchers can develop better job risk assessment to benefit industry.