ABSTRACT
This chapter reviews current thinking on appropriate analytic techniques for longitudinal data collected in a panel survey, where the initial sample design was complex in nature (possibly featuring stratification, cluster sampling, and weighting for unequal probability of selection) and attempts were made to repeatedly measure the initially sampled units at multiple follow-up waves. Many different patterns of missing data can emerge in panel surveys featuring repeated measurement of the same units over time; for example, some cases may provide complete data at every measurement occasion, others may choose not to participate at particular waves, and others may drop out of the panel survey permanently at a particular time point. This introduces the possibility of several alternative design-based and model-based approaches that simultaneously account for: the correlation of the repeated measurements on a given individual; the complex sampling features; and the different patterns of nonresponse.
