Sampling for Intensive Studies Barbara M Wildemuth and Leo L Cao
INTENSIVE SAMPLING TECHNIQUES
Most, if not all, intensive sampling involves nonprobability sampling. Since intensive studies are aimed at attempting to learn more about a particular setting or phenomenon, a nonprobabilistic approach to sampling can be a much more efficient method of focusing
130 APPLICATIONS OF SOCIAL RESEARCH METHODS
on the research questions being posed. The two approaches most appropriate for intensive studies will be discussed here: purposive sampling and theoretical sampling. (Other approaches to nonprobability sampling are briefly described in the previous chapter.)
Several common themes run through these two approaches to sampling. The first is that the quality of the sample relies on the judgment of the researcher. As you develop your sample, you will rely on your knowledge of the research question and its context as well as your knowledge of the particular setting in which you are collecting data. You will have particular reasons for including each participant, and the final sample should “present a balanced perspective” on the research questions (Rubin & Rubin, 2005, p. 64).
Purposive Sampling
Purposive sampling, as suggested by its name, is about purposively selecting specific participants for the study. Most qualitative research, especially research involving spe- cial populations, relies on the use of purposive sampling. It is intended to “maximize discovery of the heterogeneous patterns and problems that occur in the particular context under study” and to maximize “the researcher’s ability to identify emerging themes that take adequate account of contextual conditions and cultural norms” (Erlandson et al., 1993, p. 82). You will try to identify those participants who can provide you with the richest data on the phenomena in which you’re interested.
While you are not using probability sampling to draw a statistically representative sample from a population, you do need to understand the relationship between your sample and the wider universe or population to which it relates (Mason, 2002). You may be selecting a purposive sample that is, in some sense, representative of a population of interest. If so, you will want your sample to have the same characteristics (both in central tendency and variability) as the population. An alternative approach is to sample
strategically, maximizing your ability to make theoretical comparisons of interest. This
approach is usually called theoretical sampling and is discussed in the next section. A third possibility is that you want your sample to be illustrative or evocative of the phenomena of interest. Using this approach, you are likely to include a sample that vividly illustrates particular aspects of the phenomena under study.
You will also need to make a decision about the units that you will be sampling (Mason, 2002). These units may be individual people, but they may also be settings or environments, events or incidents, objects or artifacts, or texts. In many studies, you will be sampling a variety of types of units, each making a potential contribution to the study. The use of multiple types of units in combination can strengthen the validity of your findings.
Next, you will need to develop a strategy for selecting the sample. There are a number of strategies you can use. (Patton, 2002, lists 15 different strategies, only a few of which are described here.) For some studies, you may want to select “typical” cases. With this strategy, you are generally trying to argue that the elements included in your sample are similar to those not included in your sample, along a variety of dimensions that are pertinent to the study. For some studies, you may want to select extreme cases. For example, in a study of the chat reference services provided in your library, you may want to examine those online interactions that were most interactive, that is, those in which each of the participants “spoke” the most often. For some studies, you may want to select cases to maximize the variability in your sample. For instance, in a study of the handling of help desk requests, you might include some cases that were handled on first contact,
Sampling for Intensive Studies 131
some that required two to five follow-up interactions, and some that required more than five follow-up interactions. In many situations, it is important to include participants who have opposing views so that you can reach a balanced interpretation of the phenomena of interest (Rubin & Rubin, 2005). For some studies, you may want to focus on a particular type of case, minimizing the variability of your sample. For instance, you might want to examine only those help desk requests coming from humanities departments.
The purpose of your study and the research questions you are asking should guide your sampling plan. In general, you will want to select participants who are experienced with the phenomena of interest, who are knowledgeable about the research questions, and who hold various perspectives on the issues under study (Rubin & Rubin, 2005).
Theoretical Sampling
Glaser and Strauss (1967) introduced the concept of theoretical sampling as one component of their grounded theory development approach. Taking this approach, the goal of the sample is to provide data that will support the development of theory. Each element in the sample should help you “to define the properties of [your] categories; to identify the contexts in which they are relevant; to specify the conductions under which they arise, are maintained, and vary; and to discover their consequences” (Charmaz, 2000, p. 519). The focus is on the contribution to theory development that can be made by each new data source, and the simultaneous collection and analysis of data is an essential aspect of the research process. Coyne (1997) suggests that it might more accurately be called “analysis driven purposeful sampling” (p. 629) because new elements are added to the sample based on prior analysis.
“The aim of theoretical sampling is to maximize opportunities to compare events, incidents, or happenings to determine how a category varies in terms of its properties and dimensions” (Strauss & Corbin, 1998, p. 202). So how does this really work? You would initially select a setting in which to conduct your study (based on your research questions) and some initial data sources (e.g., participants, documents, artifacts, etc.). Here, at the beginning, you are using a relatively open sampling technique to broaden your data set. It is a purposive sample, as discussed previously (Coyne, 1997). Once you’ve collected some initial data, you analyze and make some initial interpretations of it. On the basis of these initial interpretations, you select additional data sources that are likely to provide comparable data. At this stage, you are trying to compare new findings, based on the new data collected, with the initial interpretations derived from your initial data collection. As Charmaz (2000) describes it, you are trying to “tease out less visible properties of [your] concepts and the conditions and limits of their applicability” (p. 519). Particular data sources might be selected because of the likelihood that they will yield data that support the examination of relationships among concepts or variations in the definitions of those concepts.
A hypothetical example may help to clarify this process. Let’s assume that you are trying to understand (i.e., generate a theory that will explain) the process through which library patrons adopt new chat reference services being offered by the library. Your initial data may be collected through interviews with a few reference librarians who provide chat reference services and a few “friends” of the library who have used the chat reference services as well as the logs of their chat sessions. Later waves of data collection may include an interview with a reference librarian hired directly from library school to get his perceptions of the training provided by the library concerning chat
132 APPLICATIONS OF SOCIAL RESEARCH METHODS
reference and how this was similar to or different than the education received while in school. You might interview a chat reference user who is from another community to understand why that person approached your library’s reference service. You might examine records from the local cable company because some of your participants have mentioned that chat would only be possible if they had access to high-speed Internet services through their cable company. Each of these data sources is different from the others, and each can provide a unique perspective on the developing theory.
Similarities and Differences Between Purposive and Theoretical Sampling
Purposive and theoretical sampling approaches are so similar that some authors use the two terms interchangeably. Others argue that theoretical sampling is a type of purposive sampling (Coyne, 1997). Both are done purposively. However, it is a requirement of theoretical sampling that it be incremental, with the sample being developed as the researcher completes the data analysis and develops the theory that will be the outcome of the study. Purposive sampling may also be conducted incrementally, but there is also the possibility that the sample can be identified at the beginning of the study and remain relatively unchanged throughout the course of the research.
SAMPLE SIZE
Intensive studies in information and library science are focused on the richness and quality of the data collected, rather than the number of study participants. While this eliminates discussion of statistical power and other factors, it does raise two related issues: how many different participants should be included in the sample, and how much data should be collected from each participant?
In situations where purposive sampling is being used, you will need to return to your research questions and the purpose of your study to guide your decisions about sample size. If a particular body of data is intended to address a particular research question, is it adequate? Have you collected enough data so that you fully understand the concept of interest and its relationship to other concepts? Have you checked for negative examples, for example, have you included data sources that are likely to contradict your preliminary theory? Think back to the original logic of your sampling plan: were you intending a representative sample, or an illustrative sample? You should be able to justify the adequacy of your sample in terms of its ability to fully address your research questions. In situations where theoretical sampling is being used, you will stop collecting addi- tional data when you have reached theoretical saturation, that is, when you’re hearing the same concepts discussed in the same way by your participants, with no additional information being added to your understanding of the theory you’re developing. Over time, the data cumulates, and later data are collected in a more focused way (based on the emerging theory) than when you started your data collection. Being able to tell when you’ve reached theoretical saturation depends on an iterative cycle of collecting and analyzing data so that as you collect additional data, you will know whether it is expanding your understanding of the phenomena of interest. If it is not, then there is no need to collect additional data from participants who are likely to repeat what you already know. The aim of this iterative process is “to refine ideas, not to increase the size of the original sample” (Charmaz, 2000, p. 519).
Sampling for Intensive Studies 133
During this process, you will want to alternate between strategies of maximizing differences among the study participants and minimizing those differences. You will begin with the minimizing strategy so that you can identify some important concepts and gain an initial understanding of them. Then you will switch to the strategy of maximizing differences so that you can spot differences in the concepts, resulting in their refinement. With this strategy, you can also find the boundaries of the phenomena or settings to which your emerging theory applies.
Whatever your sample size and whatever sampling approach you use, you need to describe your procedures and rationale in detail. Any decisions you made about who to include in the sample should be explained. You should also discuss how the selection of the sample affected the findings of the study.
ADDITIONAL ISSUES RELATED TO SAMPLING