Incomplete data: Introduction and overview
Although most longitudinal studies are designed to collect data on every individual in the sample at each time of follow-up, many studies have some missing observations. With longitudinal studies problems of missing data are far more acute than in cross-sectional studies because non-response can occur at any occasion. The term “non-response,” as used in this context, denotes that intended observations are missing. An individual’s response can be missing at one follow-up time and then be measured at a later follow-up time, resulting in a large number of distinct missingness patterns. At the same time, longitudinal studies often suﬀer from the problem of attrition or “dropout,” that is, some individuals “drop out” or withdraw from the study before its intended completion. In either case, the term “missing data” is used to indicate that intended measurements could not be obtained. Although missing data are ubiquitous in longitudinal studies when the study participants are human subjects, and much of the statistical literature has focused on these settings, similar problems with missing data also arise in longitudinal studies in the biological sciences, agriculture, and veterinary medicine.