Introduction and Preliminaries

Abstract

Missing data are a ubiquitous problem that complicate the statistical analysis of data collected in almost every discipline. For example, the analysis of change is a fundamental component of many research endeavors that employ longitudinal designs. Although most longitudinal studies are designed to collect data on every individual (or unit) in the sample at each time of follow-up, many studies have missing observations at one or more occasions. Missing data are both a common and challenging problem for longitudinal studies. Indeed, missing data can be considered the rule, not the exception, in longitudinal studies in the health sciences. Frequently, study participants do not appear for a scheduled observation, or

leave the study before its completion. When some observations are missing, the data are necessarily unbalanced over time in the sense that not all individuals have the same number of repeated measurements obtained at a common set of occasions. However, to distinguish missing data in a longitudinal study from other kinds of unbalanced data, such datasets are often referred to as being “incomplete.” This distinction is important and emphasizes the fact that intended measurements on individuals could not be obtained.