As discussed earlier in this chapter, snowball sampling is used when the members of a target population are difficult to locate. This sampling technique involves gathering data from a few members of the target population and then asking those individuals for information regarding the location of other potential members of the population. The example provided earlier in this chapter involved drug abusers. Snowball sampling is unlikely to yield a representative sample and is sometimes used at an exploratory stage of intervention development to gain initial insight into a problem area from which to design an intervention. It can also be combined with other sampling techniques. For example, in an evaluation of an intervention for family caregivers, one source of potential participants may be caregivers who attend community support groups or receive services from community agencies. After enrollment, these caregivers might be asked to recommend the program to someone else they may know in similar circumstances. This type of sampling could also be considered a form of snowballing. As mentioned previously, careful consideration must be given to human subject issues when using this type of sampling approach.
Adaptive Allocation Sampling
Adaptive allocation sampling, similar to cluster sampling, is a staged sampling approach. The initial sample is obtained using a conventional approach such as random sampling and then that sample is examined to determine if there are some geographic areas that exhibit more of the behavior/phenomena of interest on the basis of observations from a few select variables (Thompson & Collins, 2002). For example, assume a researcher is interested in evaluating an educational intervention to help remediate problems with childhood asthma. An initial random sample of households within a city is taken and prevalence of children with asthma is evaluated. In certain areas of the city, there appears to be a higher concentration of children with asthma; thus, a larger sample of households is chosen from these neighborhoods for potential inclusion in the study (for more detailed information on adaptive designs, see Thompson, 1990).