ABSTRACT
Repeated sampling of a population over time raises issues pertinent to both the sampling design and the estimation of population parameters. Regarding the former, one issue is the relationship between the sample at a future occasion to the initial sample. When sampling on just two occasions, three distinct possibilities are suggested: 1) independently selected samples, where overlap of the units selected may occur by chance; 2) the initial sample is retained in its entirety, ensuring complete overlap by design; and 3) partial overlap in sample composition which is ensured by design. When successive sampling on more than two occasions is considered, the number of possible outcomes increases. Analogous to these design issues, alternative estimators of current parameter values and their change through time are possible. In nontechnical language, these matters revolve around two questions: How much of the initial sample should be changed in succeeding samples, and in what way should previously collected data be used to estimate the value of current population parameters?
