Time is the unifying concept for a "longitudinal" or "cohort" study. The main outcomes are times to key events and/or repeated observations on a variable of interest over time. Most scientific questions are posed as regression problems in which one or more of the outcomes are expressed as functions of time and other predictor variables. We may also have interest in the joint distribution of a time-to-event and repeated measures.
Longitudinal studies make it possible to draw inferences about how processes change through time. Their analysis is more complex than for cross-sectional studies because the time-to-event outcomes are often censored and the repeated measurements are auto-correlated.
This talk will discuss the role of longitudinal studies in biomedical and public health research and the key approaches to analysis of longitudinal data. We will contrast marginal and conditional models in several situations. The ideas will be illustrated pictures and words but no equations.
|