T. Bevins, L. Duke, & S. Bevins
(1999)

Boundaries

In qualitative studies, boundary setting begins at the point of entry into the study.  This is known as gaining access.  Here one member of the group facilitates access to the rest of the group.  Gaining access refers to entering the physical location of the groups well as the level of information and personal experiences that frame the purpose of the study.  This is not always an easy process.  In ethnography, one strategy is to begin broadly and then narrow down the focus as necessary.

Sampling is the process of selecting representative units of a population for study in a research investigation.

Allows inferences and generalizations about the population without examining each unit in the population

A.  Target Population - the entire set of cases about which the researcher would like to make generalizations

B.  Accessible Population - meets the eligibility criteria and are available.  Although the intended or target population is usually evident, having access to members of this population (accessible) may be difficult.

Element - the most basic unit about which information is collected (e.g.: a subject in a study)

Eligibility Criteria - population descriptors that form the basis of selection

Representativeness - means that the characteristics of the population and the sample are congruent - that the sample is a good representation of the population

Delimitations - eligibility criteria/characteristics that restrict the population to a homogeneous group of subjects
 
 

The following are characteristic of nonprobability sampling:

1.  elements are chosen by nonrandom methods
2.  there is no way of ensuring that every element has an equal chance of being selected
3.  less rigorous, less representative

Types of nonprobability sampling:

A.  Convenience sampling:
1.  most readily accessible subjects
2.  this form of sampling has the greatest risk of bias
3.  subjects tend to be self-selecting
4.  this form of sampling is the weakest in terms of generalizability

B.  Quota Sampling:
1.  knowledge about the population of interest is used to build some representativeness into the sample
2.  identify strata and proportionally represent the size of the population strata in the sample strata
3.  addresses the problem of over/under-representatation of certain segments of the population in a sample
4.  the criterion for selection and stratification should be a variable that would reflect important differences in the dependent variable(s) under investigation
5.  still nonprobability, so an unknown source of bias may affect external validity
6.  in heterogeneous populations the risk of bias when using this form of sampling is great

C.  Purposive Sampling:
1.  knowledge about the population of interest and its elements is used to handpick the cases to be included in the sample
2.  used when a highly unusual group is being studied, or the researcher is interested in individuals who reflect different ends of the range for a particular characteristic
3.  in heterogeneous populations the risk of bias when using this form of sampling is great
4.  conscious bias in the selection of subjects remains a constant danger
5.  ability to generalize is limited since sample is handpicked
 

The following are characteristic of probability sampling:

1.  elements are chosen by random methods
2.  every element has an equal and independent chance of being selected
3.  laborious, but representative

The following are types of probability sampling:

A.  Simple random sampling:
Steps in the process of simple random sampling
1.  define population
2.  list all units in the population (a sampling frame)
3.  randomly select a sample of units from the population
Characteristics of a simple random sample:
1.  not subject to the conscious bias of the researcher
2.  representativeness is maximized
3.  differences between samples (or between the sample and the population) are purely by chance
4.  a larger sample gives more representativeness
5.  sample heterogeneity and mortality can still jeopardize representativeness
6.  this is a time-consuming and inefficient method of sampling
B.  Stratified Random Sampling:
Steps in the process of stratified random sampling
1.  the population is divided into strata (based on some characteristic like income)
2.  a number of elements from each strata are randomly selected on the basis of their proportion in the population
Characteristics of a stratified random sample:
1.  variables selected to make up the strata should be adaptable to homogeneous subsets with regard to the attributes being studied
2.  this method can increase representativeness in some situations
3.  it allows comparisons among subsets
4.  you can oversample to compensate for very small strata
5.  there may be difficulty getting a list of complete critical variable information
6.  uses time and money
 
C.  Multistage Sampling (Cluster Sampling):
1.  successive random sampling of units (units progress from large to small) (e.g.: randomly select 5 states; randomly select 5 counties in each of the chosen states; randomly select 5 cities from those counties; and randomly select your subjects from the phone books in those cities)
2.  economical
3.  has more potential for sampling errors
4.  handling the data from cluster samples is complex
 
D.  Systematic Sampling:
1.  selection of every kth case drawn from a population at fixed intervals (taking every 100th person in the phone book)
2.  to be a probability sample the sampling frame must be random in relation to the variable of interest
3.  the first element must be randomly selected
4.  more efficient
5.  bias in the form of nonrandomness can be inadvertently introduced

Matching - construct an equivalent comparison sample group with subjects who are similar to each subject in another sample group in relation to preestablished variables

Network Sampling - when a researcher has found a few subjects with the needed eligibility criteria, they are asked for their assistance in getting in touch with others with similar criteria.  DePoy refers to this as snowball sampling.
 

In quantitative research, the sample size is determined by:

• type of design
• type of sampling procedure
• a formula for estimating optimum sample size (power analysis)
• degree of precision required
• heterogeneity of attributes under investigation
• relative frequency of occurrence of the phenomenon of interest in the population
• projected cost of a particular sampling strategy

 
• In quantitative studies, the sample size should be determined before the study is conducted.  In qualitative studies sampling stops when data saturation is reached.
• Larger samples are more likely to be representative.
• Larger samples reduce the chances of type I and type II errors.
• Even a large sample cannot compensate for a faulty research design - unless representativeness is ensured, all the data in the world become inconsequential.

• should be organized
• make sure you identify the target population
• delineate the accessible population
• develop a sampling plan
• obtain approval from the Institutional Review Board

In qualitative research, we do not determine a sample size prior to conducting the data collection.  In qualitative research, which is a dynamic, fluid, flexible, inductive process, we consider what DePoy calls boundary setting.  Boundary dimensions include:

1. Geographic Location:  where the researcher plans to gather data.  This site often changes as the researcher moves into the process.  Begin by defining where the researcher enters the study.  The location may later change as the researcher gathers more information.
2. Cultural Groups:  the researcher may state that he/she is going to study certain groups in the culture.  For example, a researcher may plan to study Catholics in a poor rural area in Mexico.  The researcher may then want to know if wealth is a factor in the study , so he/she may wish to next study Catholics in a wealthy community in Mexico.
3. Personal Experience:  the boundary is set by the focus of the phenomenon being studied.  For example, a researcher may wish to study the birth families of male gang members.  Then as the study progresses, the researcher realizes he/she may need to define the boundary even more specifically - to families with a gang member in the juvenile court system.
4. Particular Concepts:  the concept may be found in the scope of the study itself.  The concept(s) of primary interest may not emerge until some initial data analysis is completed, and this emerging concept may become the focus that guides the study
5. Theory-based selection:  selecting individuals who exemplify a particular theoretical construct
6. Artifact Examined:  not discussed in DePoy.  Refers to setting the boundary of the study based on what is learned (or remains to be learned) about the artifacts being examined.
7. Research Participant Involvement: purposful action process.  Participant (subject) selections are determined from the researcher's perspective or study purpose and qualitative research question.  Selections may be formed by judgments emerging in the course of fieldwork.  The researcher selects individuals who are judged to have the potential for illuminating a particular concept, experience, or cultural context.
8. Sample size: the number of subjects and representativeness of the sample are not the focus of subject selection, but rather, subject selection depends most on the richness of information that can be obtained from any given subject.  Large ethnographic studies at time use some probability or nonprobability sampling techniques once the investigator identifies patterns that warrant these forms of subject selection.  If a homogeneous sample is used, a small number (5 to 10) may be adequate.  If maximum variation is sought in the sample, the researcher may need a sample of 20 to 50 individuals to capture all of the variation sought in the study.
9. Maximum variation v. homogeneous selection:  in qualitative research the investigator may wish to maximize variation across the broadest range of experiences, information and perspectives of study participants.  Quantitative studies are on the other hand often hampered by extreme cases (called outliers).  These outliers upset the quantitative sense of the representativeness of the sample.  Other qualitative studies, in contrast to maximum variation, may wish to have a group of subjects that are very similar.  The qualitative researcher's decision for maximum variation or homogeneous selection is based on the purpose, question and design of the study.

Confirming or disconfirming cases:  a similar concept to maximum variation or homogeneous selection is the idea that you may get to a point in your qualitative study that you search for subjects that either support or challenge an emerging theory.  Understanding a concept or developing a theory can be supported by either of these strategies, and selecting one or the other again depends on the purpose, question and design of the study, as well as an analysis of the data collected up to that point.  The researcher would need to decide whether the concept or theory is better understood or developed by additional support of similar information or the additional information gained by looking in a different direction or more extreme cases.