Nonprobability Sampling Techniques
Probability sampling methods are not always practical and are not typically used in intervention research. Rather, nonprobability sampling techniques are commonly used. Using these sampling methods, it is not possible to specify the probability or likelihood of specifying and then selecting one individual over another. In other words, in behavioral intervention research, not everyone in the target population has an equal chance of being included in the sample. Thus, the sample may not be representative of the population, which threatens the external validity of the study or the extent to which one can generalize intervention outcomes from the sample to the population. For this reason, it is important to clearly specify the key characteristics of the population being targeted by the intervention and then to select a sample as representative of that population as possible. For example, assume a researcher is interested in evaluating the efficacy of a computer-training program for seniors with low computer literacy; it would be important to select older adults of both genders, mixed ethnic/racial backgrounds, of a fairly broad age range (e.g., 65+ years), with limited computer skills. Individuals with strong computer skills in the training program should be excluded from the study. If they were to be included, it would limit the extent to which one could generalize findings to the main population of interest. Furthermore, those with high skill levels may rate the class as too simplistic or slow in pace. As noted in Chapter 10, it is important to clearly specify criteria for including and excluding individuals early on in the development of the intervention. Identifying who will most likely benefit from an intervention and who will not is part of the initial work of an interventionist in the discovery phase of the pipeline (Chapter 3).
The most common methods of nonprobability sampling are convenience sampling, quota sampling, purposive sampling, snowballing techniques, and adaptive allocation. Each of these techniques will be described in the following sections. However, before beginning a review of these techniques, we will begin this section with a brief discussion of adaptive sampling since adaptive sampling techniques (e.g., snowball sampling and adaptive allocation sampling, which are described later) are often used when conducting research with populations such as people at high risk for infection, substance abusers, or people who are homeless or severely mentally ill. Adaptive sampling approaches involve using information gained during initial sampling of participants to redirect sampling strategies (Thompson & Collins, 2002). In the substance abuse example provided earlier, an initial sample of drug users might be asked for names of other users with whom they interact. These individuals are then approached to determine if they would be willing to participate in the study (of course, they would not be enrolled without consent). In this case, additional study participants are added on the basis of social network information obtained through contact with initial members of the sample as opposed to the information established prior to the beginning of recruitment. In other words, information gathered during initial sampling is used to inform future sample efforts. Adaptive sampling designs generally involve those that are based on social networks (e.g., snowballing) and those that are based on geographic location (adaptive allocation design). We provide a brief overview of these techniques later. However, as noted by Thompson and Collins (2002), statistical theory and procedures for estimating population quantities for these designs are still evolving (see Thompson & Collins, 2002 for a more complete discussion of these issues). It is also important to give careful consideration to ethics and human subject issues (see Chapter 13). For example, someone may be uncomfortable providing the name of someone else who is also involved in illicit behaviors. Overall, adaptive designs offer many advantages when dealing with challenging health and human service issues in which common approaches to reaching out to the targeted population do not suffice. However, it is important to be aware of the limitations of these approaches, especially with respect to their implications for external validity.