Question

Describe the differences in sampling techniques used in quantitative versus qualitative research. How does a researcher know when they have "enough" data when using each methodology?

Answer 1

Sampling and types of sampling methods commonly used in quantitative research are discussed in the following module.

Learning Objectives:

• Define sampling and randomization.
• Explain probability and non-probability sampling and describes the different types of each.

Researchers commonly examine traits or characteristics (parameters) of populations in their studies. A population is a group of individual units with some commonality. For example, a researcher may want to study characteristics of female smokers in the United States. This would be the population being analyzed in the study, but it would be impossible to collect information from all female smokers in the U.S. Therefore, the researcher would select individuals from which to collect the data. This is called sampling. The group from which the data is drawn is a representative sample of the population the results of the study can be generalized to the population as a whole.

The sample will be representative of the population if the researcher uses a random selection procedure to choose participants. The group of units or individuals who have a legitimate chance of being selected are sometimes referred to as the sampling frame. If a researcher studied developmental milestones of preschool children and target licensed preschools to collect the data, the sampling frame would be all preschool aged children in those preschools. Students in those preschools could then be selected at random through a systematic method to participate in the study. This does, however, lead to a discussion of biases in research. For example, low-income children may be less likely to be enrolled in preschool and therefore, may be excluded from the study. Extra care has to be taken to control biases when determining sampling techniques.

There are two main types of sampling: probability and non-probability sampling. The difference between the two types is whether or not the sampling selection involves randomization. Randomization occurs when all members of the sampling frame have an equal opportunity of being selected for the study. Following is a discussion of probability and non-probability sampling and the different types of each.

Probability Sampling – Uses randomization and takes steps to ensure all members of a population have a chance of being selected. There are several variations on this type of sampling and following is a list of ways probability sampling may occur:

• Random sampling – every member has an equal chance
• Stratified sampling – population divided into subgroups (strata) and members are randomly selected from each group
• Systematic sampling – uses a specific system to select members such as every 10^th person on an alphabetized list
• Cluster random sampling – divides the population into clusters, clusters are randomly selected and all members of the cluster selected are sampled
• Multi-stage random sampling – a combination of one or more of the above methods

Non-probability Sampling – Does not rely on the use of randomization techniques to select members. This is typically done in studies where randomization is not possible in order to obtain a representative sample. Bias is more of a concern with this type of sampling. The different types of non-probability sampling are as follows:

• Convenience or accidental sampling – members or units are selected based on availability
• Purposive sampling – members of a particular group are purposefully sought after
• Modal instance sampling – members or units are the most common within a defined group and therefore are sought after
• Expert sampling – members considered to be of high quality are chosen for participation
• Proportional and non-proportional quota sampling – members are sampled until exact proportions of certain types of data are obtained or until sufficient data in different categories is collected
• Diversity sampling – members are selected intentionally across the possible types of responses to capture all possibilities
• Snowball sampling – members are sampled and then asked to help identify other members to sample and this process continues until enough samples are collected

• Sampling for Qualitative Research

  Sampling, as it relates to research, refers to the selection of individuals, units, and/or settings to be studied. Whereas quantitative studies strive for random sampling, qualitative studies often use purposeful or criterion-based sampling, that is, a sample that has the characteristics relevant to the research question(s). For example, if you are interested in studying adult survivors of childhood sexual abuse, interviewing a random sample of 10 people may yield only one adult survivor, thus, you will essentially have a sample size of one and need to continue to randomly sample people until you have interviewed an appropriate number of who have survived childhood sexual abuse. This is not a wise use of your time.

  The difference in sampling strategies between quantitative and qualitative studies is due to the different goals of each research approach. Recall that typical quantitative research seeks to infer from a sample to a population (for example, a relationship or a treatment effect). In general, you want to include a variety of types of people in a quantitative study so that it generalizes beyond those in your study. Thus, the goal of quantitative approaches can be stated as, ”empirical generalization to many.”

  Qualitative research, on the other hand, typically starts with a specific group, type of individual, event, or process. As in the qualitative study of adult survivors of childhood sexual abuse example above, you would choose your sample very purposefully and include in your study only those with this particular experience. The goal of qualitative research can be stated as “in-depth understanding.”

  It is true that some aspects of quantitative sampling could be relevant to a qualitative researcher. For example, if you are interested in children’s experiences of Hurricane Katrina and you have access to 3,000 school children, all of whom experienced the hurricane, you might choose to randomly sample 10 children from the 3,000 for your qualitative study. In the case of ethnographic survey research, you might even seek to obtain sample sizes similar to those in a quantitative design. It could be said, then, that there are more ambiguities than “rules” when it comes to qualitative research in general and that choosing a sampling strategy and sample size for qualitative research is no different. What is important to remember is that the strategy you adopt will be driven by the:

• Research question(s)/purpose
• Time frame of your study
• Resources available

• Following is a list of common sampling strategies. As you read these strategies, think of which would be most relevant for your area of interest. In many cases, you will see ways to combine the strategies to create an effective approach. For example, you may use snowball sampling as a method to identify a set of extreme/deviant cases. This is an example of combination or mixed purposeful sampling. Thus these methods are not mutually exclusive; a research design may adopt a range of strategies.

  Common Qualitative Sampling Strategies [1]

• Extreme or Deviant Case Sampling—Looks at highly unusual manifestations of the phenomenon of interest, such as outstanding success/notable failures, top of the class/dropouts, exotic events, crises. This strategy tries to select particular cases that would glean the most information, given the research question. One example of an extreme/deviant case related to battered women would be battered women who kill their abusers.
• Intensity Sampling—Chooses information-rich cases that manifest the phenomenon intensely, but not extremely, such as good students/poor students, above average/below average. This strategy is very similar to extreme/deviant case sampling as it uses the same logic. The difference is that the cases selected are not as extreme. This type of sampling requires that you have prior information on the variation of the phenomena under study so that you can choose intense, although not extreme, examples. For example, heuristic research uses the intense, personal experience(s) of the researcher. If one were studying jealousy, you would need to have had an intense experience with this particular emotion; a mild or pathologically extreme experience would not likely elucidate the phenomena in the same way as an intense experience.
• Maximum Variation Sampling—Selects a wide range of variation on dimensions of interest. The purpose is to discover/uncover central themes, core elements, and/or shared dimensions that cut across a diverse sample while at the same time offering the opportunity to document unique or diverse variations. For example, to implement this strategy, you might create a matrix (of communities, people, etc.) where each item on the matrix is as different (on relevant dimensions) as possible from all other items.
• Homogeneous Sampling—Brings together people of similar backgrounds and experiences. It reduces variation, simplifies analysis, and facilitates group interviewing. This strategy is used most often when conducting focus groups. For example, if you are studying participation in a parenting program, you might sample all single-parent, female head of households.
• Typical Case Sampling—Focuses on what is typical, normal, and/or average. This strategy may be adopted when one needs to present a qualitative profile of one or more typical cases. When using this strategy you must have a broad consensus about what is “average.” For example, if you were working to begin development projects in Third World countries, you might conduct a typical case sampling of “average” villages. Such a study would uncover critical issues to be addressed for most villages by looking at the ones you sampled.
• Critical Case Sampling—Looks at cases that will produce critical information. In order to use this method, you must know what constitutes a critical case. This method permits logical generalization and maximum application of information to other cases because if it's true of this one case, it's likely to be true of all other case. For example, if you want to know if people understand a particular set of federal regulations, you may present the regulations to a group of highly educated people (“If they can’t understand them, then most people probably cannot”) and/or you might present them to a group of under-educated people (“If they can understand them, then most people probably can”).
• Snowball or Chain Sampling—Identifies cases of interest from people who know people who know what cases are information-rich, that is, who would be a good interview participant. Thus, this is an approach used for locating information-rich cases. You would begin by asking relevant people something like: “Who knows a lot about ___?” For example, you would ask for nominations, until the nominations snowball, getting bigger and bigger. Eventually, there should be a few key names that are mentioned repeatedly.
• Criterion Sampling—Selects all cases that meet some criterion. This strategy is typically applied when considering quality assurance issues. In essence, you choose cases that are information-rich and that might reveal a major system weakness that could be improved. For example, if the average length of stay for a certain surgical procedure is three days, you might set a criterion for being in the study as anyone whose stay exceeded three days. Interviewing these cases may offer information related to aspects of the process/system that could be improved.
• Theory-Based or Operational Construct or Theoretical Sampling—dentifies manifestations of a theoretical construct of interest so as to elaborate and examine the construct. This strategy is similar to criterion sampling, except it is more conceptually focused. This strategy is used in grounded theory studies. You would sample people/incidents, etc., based on whether or not they manifest/represent an important theoretical or operational construct. For example, if you were interested in studying the theory of “resiliency” in adults who were physically abused as children, you would sample people who meet theory-driven criteria for “resiliency.”
• Confirming and Disconfirming Sampling—Seeks cases that are both “expected” and the “exception” to what is expected. In this way, this strategy deepens initial analysis, seeks exceptions, and tests variation. In this strategy you find both confirming cases (those that add depth, richness, credibility) as well as disconfirming cases (example that do not fit and are the source of rival interpretations). This strategy is typically adopted after initial fieldwork has established what a confirming case would be. For example, if you are studying certain negative academic outcomes related to environmental factors, like low SES, low parental involvement, high teacher to student ratios, lack of funding for a school, etc. you would look for both confirming cases (cases that evidence the negative impact of these factors on academic performance) and disconfirming cases (cases where there is no apparent negative association between these factors and academic performance).
• Stratified Purposeful Sampling—Focuses on characteristics of particular subgroups of interest; facilitates comparisons. This strategy is similar to stratified random sampling (samples are taken within samples), except the sample size is typically much smaller. In stratified sampling you “stratify” a sample based on a characteristic. Thus, if you are studying academic performance, you would sample a group of below average performers, average performers, and above average performers. The main goal of this strategy is to capture major variations (although common themes may emerge).
• Opportunistic or Emergent Sampling—Follows new leads during fieldwork, takes advantage of the unexpected, and is flexible. This strategy takes advantage of whatever unfolds as it is unfolding, and may be used after fieldwork has begun and as a researcher becomes open to sampling a group or person they may not have initially planned to interview. For example, you might be studying 6th grade students’ awareness of a topic and realize you will gain additional understanding by including 5^th grade students’ as well.
• Purposeful Random Sampling—Looks at a random sample. This strategy adds credibility to a sample when the potential purposeful sample is larger than one can handle. While this is a type of random sampling, it uses small sample sizes, thus the goal is credibility, not representativeness or the ability to generalize. For example, if you want to study clients at a drug rehabilitation program, you may randomly select 10 of 300 current cases to follow. This reduces judgment within a purposeful category, because the cases are picked randomly and without regard to the program outcome.
• Sampling Politically Important Cases—Seeks cases that will increase the usefulness and relevance of information gained based on the politics of the moment. This strategy attracts attention to the study (or avoids attracting undesired attention by purposefully eliminating from the sample politically sensitive cases). This strategy is a variation on critical case sampling. For example, when studying voter behavior, one might choose the 2000 election, not only because it would provide insight, but also because it would likely attract attention.
• Convenience Sampling—Selects cases based on ease of accessibility. This strategy saves time, money, and effort, however, has the weakest rationale along with the lowest credibility. This strategy may yield information-poor cases because cases are picked simply because they are easy to access, rather than on a specific strategy/rationale. Sampling your co-workers, family members or neighbors simply because they are “there” is an example of convenience sampling.
• Combination or Mixed Purposeful Sampling—Combines two or more strategies listed above. Basically, using more than one strategy above is considered combination or mixed purposeful sampling. This type of sampling meets multiple interests and needs. For example, you might use chain sampling in order to identify extreme or deviant cases. That is, you might ask people to identify cases that would be considered extreme/deviant and do this until you have consensus on a set of cases that you would sample.