ABSTRACT
Data are de˜ned as being longitudinal if each individual, or other experimental unit, has more than one measurement taken through time. This contrasts with what are sometimes called cross-sectional data, which consist of a single outcome on each experimental unit. It is only when a study is designed to obtain longitudinal data that an assessment of within-unit changes in the response over time can be made. In the context of population studies, longitudinal data are required in order to be able to separate out what are termed “cohort” and “age” effects. Cohort effects are caused by differences between the groups of individuals in the study, typically with each group being a different age cohort. Age effects are caused by changes that occur within an individual over time. In a purely cross-sectional study, when only one measurement is taken from each individual, even when the individuals span a wide range with respect to their age, the cohort and age effects are completely confounded.
