ABSTRACT
A fairly common challenge in using samples to infer properties of populations is that some of the population may be inaccessible or “closed-off” to sampling. For instance, say the population of interest is all adults in a geographic jurisdiction, but the sampling proceeds via telephone calls to “landline” phone numbers. Then the portion of the population not having landlines will be inaccessible. The challenge can become particularly acute when dealing with socalled “hard-to-reach” populations, for which most forms of probability-based sampling are difficult, if not impossible, to implement. Examples of hard-toreach populations include populations defined by illicit drug use, and populations defined by sexual behaviors. There is considerable recent literature on both sampling hard-to-reach populations and appropriately analyzing the resulting data. Some prominent sampling schemes include “respondent-driven sampling,” “snowball sampling,” and “venue sampling.” As a few recent entry points into this literature, see Gile and Handcock (2010), Johnston and Sabin (2010), Semaan (2010), Karon and Wejnert (2012), and Gustafson et al. (2013).
