A review of methods for modelling both Gaussian and Non-Gaussian longitudinal data with application.

Abstract

The study of longitudinal data plays an integral role in medicine, epidemiology, social science, biomedical and health sciences research where repeated measurements are obtained over time for each individual. Generally, the interest is in the dependence of the outcome variable on the covariates. The analysis of the data from longitudinal studies requires special techniques, which take into account the fact that the repeated measurements within one individual are correlated. In review of this work, we explore modern developments in the area of linear and nonlinear generalized mixed-effects regression models and various alternatives including generalized estimating equations for analysis of longitudinal data and correspondence analysis (CA). Methods are described for continuous and normally distributed as well as categorical variables. We apply this theory to the analysis of complete longitudinal data from National Institute of Environment Health Sciences (NIEHS) focusing on the relationship between blood lead levels (PbB) and some associated covariates. The results show that Placebo-treated children had a gradual (occuring) decrease in blood lead level. Succimer-treated children had an abrupt (unexpected) drop in blood lead level, followed by rebound. The average mean blood lead level of the succimer-treated children after initiation of treatment was 19.14 μg/dL lower than that of placebo-treated children. After randomization, blood lead levels had fallen by similar amounts in both chelated and placebo children, despite the immediate drops in the chelated group; there was no association between change in blood lead level and change in cognitive test score. Blood lead levels continued to fall.