Recent Developments of Sampling Hard-to-Reach Populations: An AssessmentView Presentation
*Sunghee Lee, University of Michigan
Keywords: sampling; nonprobability sampling
The ever-increasing needs for studying hard-to-reach populations present a set of new challenges to samplers. Unlike data collection with the general population where probability sampling has provided a theoretical basis for inference to the population, probability sampling for hard-to-reach populations is often deemed infeasible due to prohibitively high costs and a sensitive nature of some populations. Sampling statistics as a field has long acknowledged such challenges but has yet to provide practical solutions. Meanwhile, the demands for data on hard-to-reach populations seem to have driven methodological considerations in other fields much further than in sampling statistics. Public health, in particular, has seen various sampling methods suggested for hard-to-reach populations. New sampling methods, such as respondent-driven sampling, community-based sampling, and time-and-space sampling, are easily located in the recent public health literature where these methods appear to have been accepted as standard approaches. Despite their popularity, these new sampling techniques are not found in traditional sampling textbooks as they are mostly based on non-probability methods. This paper attempts to provide an overview of these methods for researchers interested in investigating hard-to-reach populations. First, we will introduce each sampling method with its theoretical background and critical assumptions required for inferences. Second, we will examine the characteristics of hard-to-reach populations sampled with these methods and specific applications in published studies. Third, we will illustrate when the assumptions are likely and unlikely to be met in data collection operations. Last, the benefits and limitations of these methods will be discussed from a total survey error perspective.