IJDSv16p339 362casteel7067
IJDSv16p339 362casteel7067
ABSTRACT
Aim/Purpose The purpose of this article is to present clear definitions of the population
structures essential to research, to provide examples of how these structures are
described within research, and to propose a basic structure that novice research-
ers may use to ensure a clearly and completely defined population of interest
and sample from which they will collect data.
Background Novice researchers, especially doctoral students, experience challenges when de-
scribing and distinguishing between populations and samples. Clearly defining
and describing research structural elements, to include populations and the sam-
ple, provides needed scaffolding to doctoral students.
Methodology The systematic review of 65 empirical research articles and research texts pro-
vided peer-reviewed support for presenting consistent population- and sample-
related definitions and exemplars.
Contribution This article provides clear definitions of the population structures essential to
research, with examples of how these structures, beginning with the unit of
analysis, are described within research. With this defined, we examine the popu-
lation subsets and what characterizes them. The proposed writing structure pro-
vides doctoral students a model for developing the relevant population and
sample descriptions in their dissertations and other research.
Findings The article describes that although many definitions and uses are relatively con-
sistent within the literature, there are epistemological differences between re-
search designs that do not allow for a one-size-fits-all definition for all terms.
We provide methods for defining populations and the sample, selecting a sam-
ple from the population, and the arguments for and against each of the meth-
ods.
Accepting Editor Eli Cohen │Received: January 22, 2021│ Revised: April 29, 2021 │ Accepted: May 3, 2021.
Cite as: Casteel, A., & Bridier, N. L. (2021). Describing populations and samples in doctoral student research.
International Journal of Doctoral Studies, 16, 339-362. https://doi.org/10.28945/4766
(CC BY-NC 4.0) This article is licensed to you under a Creative Commons Attribution-NonCommercial 4.0 International
License. When you copy and redistribute this paper in full or in part, you need to provide proper attribution to it to ensure
that others can later locate this work (and to ensure that others do not accuse you of plagiarism). You may (and we encour-
age you to) adapt, remix, transform, and build upon the material for any non-commercial purposes. This license does not
permit you to use this material for commercial purposes.
Describing Populations and Samples
Recommendations Social science research faculty seek structured ways in which to present key re-
for Practitioners search elements to doctoral students and to provide a model by which they may
write the dissertation. The article offers contemporary examples from the peer-
reviewed literature to support these aims.
Recommendations Novice researchers may wish to use the recommended framework within this
for Researchers article when developing the relevant section of the dissertation. Doing so pro-
vides an itemized checklist of writing descriptions, ensuring a more complete
and comprehensive description of the study population and sample.
Impact on Society The scientific method provides a consistent methodological approach to re-
searching and presenting research. By reemphasizing the definitions and appli-
cations of populations and samples in research, and by providing a writing
structure that doctoral students may model in their own writing, the article sup-
ports doctoral students’ growth and development in using the scientific method.
Future Research Future researchers may wish to further advance novice researcher knowledge in
developing models to guide dissertation writing. Future studies may focus on
other essential areas of research, including studies about recruitment methods
and attrition strategies, data collection procedures, and overall research align-
ment. Additionally, future researchers may wish to consider evaluating doctoral
student foundational knowledge about populations and samples as part of the
research process.
Keywords population of interest, target population, sampling frame, sample, unit of analy-
sis, unit of observation
INTRODUCTION
As novice researchers, doctoral students are new to the processes, procedures, and philosophical
principles of developing research. Doctoral students experience challenges including, but not limited
to, difficulties in selecting a research topic (Simui et al., 2018), recruiting participants (Marks et al.,
2017), developing interview protocols (Wohlfart, 2020), and data analysis (Lewis, 2020). An additional
challenge is adequately describing the population and sample of the study. To generalize results, one
must be able to infer or transfer those results to the appropriate group. However, the grouping of
individuals can be complex and unpredictable, individuals may or may not want to be scrutinized and
the characteristics being examined may change from day to day or even from hour to hour (Ham-
mersley & Mairs, 2004). Human characteristics are often hidden from superficial examination, pre-
senting a challenge for the researcher in selecting an appropriate group to study. As such, researchers
are challenged with the task of identifying the characteristics of groups to be studied, and then de-
scribing the boundaries – the features that determine one’s inclusion or exclusion to that group – and
doing so in a relevant context (Porzsolt et al., 2019).
The challenge of defining and describing populations and samples becomes even more confusing to
novice researchers. The purpose of this article is to present clear definitions of the population struc-
tures essential to research, to provide examples of how these structures are described within the liter-
ature, and to propose a basic structure that novice researchers may use to ensure they have clearly
and completely defined the population they wish to study and the sample from which they will gather
data. In this article, we begin with the fundamental building block of any population: the unit of
analysis. With this defined, we examine the population subsets and what characterizes them. We then
examine methods for selecting a sample from the population and the arguments for and against each
of the methods. Finally, we propose a fundamental outline to aid novice researchers in developing
and describing the individuals, groups, or organizations being studied. We use Bakibinga et al.’s (2019)
research of slums across four countries to illustrate the approaches used to describe various popula-
tions and sampling methods discussed within this paper.
340
Casteel & Bridier
UNIT OF ANALYSIS
Social sciences often explore individual characteristics, such as psychological constructs, perceptions,
and behaviors. Derived from the appropriate theoretical framework or model, the data operationalize
those constructs, perceptions, and behaviors, as measured by instruments or via symbol-based means.
The data are then used to describe the unit of analysis, which in social sciences is often the individ-
ual. Throughout a given research process, the unit of analysis – the individual – is the entity being
examined and ultimately analyzed to provide a conclusion that explains the outcome and addresses
the research problem. The individual described or measured by the study variables represents the unit
of analysis (Salkind, 2010). More specifically, the unit of analysis is the entity described by the data col-
lected and about which the analysis is conducted. It is what the data are about for the purpose of ad-
dressing the research problem and defines the boundaries of what is examined or ignored within the
study (Ritella et al., 2020). The importance of clearly understanding and defining the unit of analysis
cannot be overstated. The unit of analysis has been written about as the “most important part of any
research…as the whole research is based upon the unit of analysis” (Khan, 2014, p. 228), and as the
“key to developing concepts, empirically measuring or observing concepts, and using data analysis”
(Neuman, 2013, p. 69).
For example, one may examine how student characteristics – whether student performance (e.g.,
grades, test scores), individual constructs (e.g., self-efficacy, personality traits, motivation), or behav-
iors (e.g., responses to stimuli, disruptive conduct) – may compare from one environment to the
next. In such cases, the measures apply to the individual student, and the results are applied to indi-
viduals. The individual student is the unit of analysis with all data describing the individual student
characteristics. It should also be no surprise that the theoretical frameworks supporting these types
of investigations are oriented toward individual characteristics (e.g., Bandura’s [1977] social learning
theory, McCrae and Costa’s [2003] five-factor theory of personality traits, and Deci and Ryan’s [2008]
self-determination theory).
In contrast, if one is comparing organizational cultures between different companies, one would ag-
gregate the measures from throughout each company, and then compare the composite measures be-
tween organizations (Kumar, 2018). One would examine the entirety of organizational behaviors
when examining activities, such as mergers and acquisitions. In this case, the organization is the unit
of analysis with the data describing characteristics of the organization, which is then used in the anal-
ysis. Consequently, the results describe organizational differences or relationships. The theoretical
framework from which the variables are developed for such a study are oriented toward organiza-
tional characteristics (e.g., transaction cost economics [Coase, 1937], and organizational commitment
theory [Meyer & Allen, 1997]). Regardless of what constitutes the unit of analysis, whether using dy-
ads (e.g., father-child relationships [Kenny et al., 2006]), schools, or politically bounded units (e.g.,
municipalities, county, state, providence, or national governments), the data are collected, and the
study variables apply to the entity being examined. Consequently, the results are then applicable to
the broader population of the same units of analysis (Campbell, 1986).
In qualitative designs, the data are collected to explore the unit of analysis, which may be a phenome-
non, lived experience (Dieumegard et al., 2019), case, individual, or group (Yin, 2014), depending
upon the nature of the research design and of the research problem. Although one must allow for
researcher flexibility in determining the unit of analysis based upon the research problem, there are
some examples from the literature modeling the use of units of analysis within qualitative research.
Recalling that the unit of analysis is the unit about which data are collected, Colorafi and Evans
(2016) describe the data as being applied toward understanding the phenomenon of interest, while
Magilvy and Thomas (2009) describe the data analysis for making sense and understanding the phe-
nomenon of the study. In each case, the phenomenon is what the data are about within these de-
scriptions of qualitative descriptive research designs.
341
Describing Populations and Samples
Within case study research, the unit of analysis is regularly described as the case with the phenome-
non occurring within the bounded context of the case (Baxter & Jack, 2008). Case studies may exam-
ine individuals, groups, processes, events, decisions, programs, or any other phenomena that are en-
cased by temporal and spatial boundaries (Yin, 2014). The boundaries of the case place necessary re-
strictions on what data will be collected and the context in which the data will be analyzed to prevent
too broad of an analysis and too unwieldy a research project. Based upon the working definition that
the unit of analysis is what the data are about, the case certainly qualifies to be characterized as the
unit of analysis. It is worthy of note that not all authors agree that the case is the unit of analysis, in-
stead further defining the unit of analysis as being embedded within the case layers as to be the event
of interest or phenomenon about which the case is centered, with the potential for multiple units of
analysis within an embedded case design (Grünbaum, 2007).
Giorgi (2006) describes the theme, or central focus, of the descriptive phenomenological investiga-
tion as the object – the event that is taken up by the experiencer, which may also be described as the
lived experience (Dieumegard et al., 2019). Within ethnographic studies, the unit of analysis is regu-
larly defined as the ethnos, a description of the nation, tribe, or people (Erickson, 1984). Erickson
goes on to argue that the ethnos may extend to any social network, necessitating a clear definition of
the network’s boundaries, just as one must describe the limitations of the nation, tribe, or group of
people being studied. Within narrative research design, the discrete story is the unit of analysis, in
which it is data that describe the sequence and consequence of events that are analyzed and shared
with the reader (Riessman & Quinney, 2005). Grounded theory research design is used to explore the
process, action, or interaction between many individuals in which the process, action, or interaction is
the unit of analysis (Khan, 2014). Regardless of the research design, the definition of the unit of
analysis must be clearly delineated so the researcher and participants understand who or what is in-
cluded within the unit and who or what is excluded from the unit to understand what the research
problem is addressing.
S LUMS E XAMPLE
Bakibinga et al. (2019) provides an excellent example of population and sample descriptions within
the research, beginning with a description of the unit of analysis. In preparation of a quantitative
study, Bakibinga et al. identify the need to collect data about the residents of slums across the globe
and set about describing the process by which this was achieved. A first order of business was pre-
cisely describing the unit of analysis, which was the household within the slum. The research prob-
lem dictated the data were to be collected about the individual household and the analyses were to
draw conclusions about the households within slums. The authors then defined what constitutes the
unit of analysis, the household, which was described as the residents of a residential dwelling within
a slum, as identified by the head of household’s name.
From the examples provided, it is clear that identifying and describing the unit of analysis is one of
the first steps of any research study. Together with clearly defined variables or phenomena of inter-
est, the unit of analysis helps build the foundation for ensuring the selection of a sample most repre-
sentative of the population of interest to adequately address the research problem. Since the unit of
analysis is the entity about which all variables apply and the entity that is described by the data col-
lected, we must also consider the source of the data collected.
UNIT OF OBSERVATION
While the unit of analysis is the entity about which the data are reported, the unit of observation is
the entity that is the source of data about the unit of analysis. The unit of observation may include
the participant, who is self-reporting, in which case the unit of observation and unit of analysis are
the same (Kumar, 2018). However, these are not always the circumstances. The unit of observation
may include a parent who provides data about their child (unit of analysis), an employee (unit of ob-
servation) reporting characteristics of a manager (unit of analysis), a primary instrument, such as
342
Casteel & Bridier
POPULATION OF INTEREST
The population of interest for the study is comprised of the individuals, dyads, groups, organiza-
tions, or other entities one seeks to understand and to whom or to which the study results may be
generalized or transferred and is the principal group about which the research is concerned. Popula-
tions create boundaries for the scope of a study and provide environmental and context cues for the
reader. Such boundaries place natural delimitations upon the research to afford the researcher the
proper focus so as not to present a one-size-fits-all set of results. The definition of boundaries also
allows the researcher to clearly identify subpopulations, such as the target population, sampling
frame, and sample, and to ensure alignment between these groups within the research (Salkind,
2010).
Clear definitions of boundaries and other considerations are essential to describing each respective
group. It is incumbent upon the researchers to define the boundaries used in their study to avoid
compelling the reader to assume a definition; otherwise, the incorrect definition – and, therefore, the
incorrect population – may be presumed by the reader. Additionally, boundaries described by social
or political constructs may vary by geographical area (Bakibinga et al., 2019) or by researcher or theo-
retical perspective. Such populations may be heterogeneous, requiring further identification and de-
scription of the population to be studied. Needless to say, a rich and thorough description of each
group to be studied is required. Defining the population of interest provides context for who may be
characterized by and interested in the results of the study. Providing a rich description of the popula-
tion of interest allows the community of readers and practitioners to associate the research with the
appropriate entities, whether they are people or organizations. The overall results are that the popula-
tion becomes better informed about itself.
Boundaries for the population of interest may include geography, race/ethnicity, age, gender, sexual
orientation, marital status, education, income, household composition, occupation, business sector,
number of employees, school district, dyadic relationship type, government-defined, and so forth
(Sudman, 1976). The boundaries are developed such that the variables or experience of interest are
operationalized within those boundaries. To operationalize the boundaries for the population of in-
terest, one should choose characteristics that are easily identifiable or recognizable and that are asso-
ciated with the variables or experiences of interest. Doing so may include some individuals or entities
who do not exhibit the characteristics in which one is interested and may exclude others who do ex-
hibit those characteristics, but, in general, the population represents the individuals to whom the re-
sults apply and matter.
It is easy to swing the pendulum too far to one side when describing the population. Caution is rec-
ommended to avoid over-defining the population of interest, as it (often unnecessarily) restricts the
applicability of the study and the ability to obtain a sample, actions that increase the time required to
complete the study, as well as the costs to do so. When examining general psychological constructs,
such as personality traits, selecting, describing, and examining broad populations seem naturally
343
Describing Populations and Samples
appropriate, such as Soto and John (2017), whose updated Big Five Inventory (BFI-2) was tested (N
= 1,137) to measure personality traits and facets for Western, English-speaking adults.
However, for studies of narrower focus, such as Ojala et al. (2014), a study conducted using a phe-
nomenological method to study of the effects of chronic pain upon the whole person (N = 34), the
population of interest is reduced somewhat. Although the study was conducted in Finland, the gen-
eral results from the patients’ experience of the phenomenon are transferrable to an international au-
dience of chronic pain sufferers, which was the population of interest.
TARGET POPULATION
The target population is the specific, conceptually bounded group of potential participants to whom
the researcher may have access that represents the nature of the population of interest. To be suc-
cessful in defining the target population, one must examine all the boundary considerations in an iter-
ative manner to ensure that the end description of the target population is inclusive enough to pro-
vide sufficient data to the study. The target population must also be exclusive enough to avoid having
participants who do not represent the study’s needs, which will misrepresent the population of inter-
est. Much like the population of interest, the boundaries of the target population must be defined
such that the researcher and other stakeholders understand the nature and extent of the group to be
studied. Such considerations are important not only for ensuring the efficacy of the research, but also
assist in budgeting resources for investigating the research problem. A well-defined target population
describes inclusion and/or exclusion criteria for who or for which entities may participate in the
study. The target population must be a complete subset of the population of interest – members of
the target population must also be described by the boundaries of the population of interest. Addi-
tionally, the target population is further restricted such that the researcher may clearly operationalize
the boundaries for participation (Kalleberg et al., 1990). It is from the target population that the sam-
pling frame is developed.
To determine the target population, one must operationalize the unit’s characteristics of interest
based upon the study variables or the qualitative experience so the results of the study may be accu-
rately inferred or transferred back to the population of interest (Ackerman et al., 2019). Operational-
izing the characteristics also benefits both the researcher and potential participants so they may rec-
ognize each participant’s eligibility for the study. Very often the initial characteristics are relatively easy
to apply, and this description gives the researcher insight into the target population and the popula-
tion of interest, as a whole. However, more detail is required to fully define and operationalize the
boundaries for the target population.
A critical step in operationalizing the criteria is clearly defining the key terms and ensuring the defin-
ing words are unambiguous. For example, Gross et al. (2018), who examined approaches to mental
health and sport performance in female collegiate athletes, were very clear on the boundaries for in-
clusion within the study. The authors clearly defined and limited participation to current National
Collegiate Athletic Association Division III female athletes.
344
Casteel & Bridier
Additionally, the target population must incorporate an additional set of boundaries that include time
and space to anchor and identify the group of individuals who will be asked to participate in the
study. The temporal element restricts the timeframe that the participants have exhibited the opera-
tionalized inclusion variables – many professional athletes were once collegiate athletes but should
not be included in the Gross et al. (2018) example. A difference that distinguishes the two is time. By
stating that the potential participant must currently be a member of intercollegiate athletic squad, one
establishes the time-based restriction that is necessary to operationalize the sampling frame bounda-
ries. The temporal framing also eliminates individuals who are still in college but no longer play for
an intercollegiate team, and who are not currently experiencing the phenomenon being examined.
The spatial boundary is often a matter of research logistics. Although it would seem desirable to cre-
ate a target population that includes all female intercollegiate athletes across the nation’s universities,
it is not logistically practical to do so. One must place spatial restrictions on the target population to
accommodate responsible research. The restrictions often take the form of geographical descrip-
tions, such as within the United States, Australia, or United Kingdom, within a specific pre-defined
sports conference, or more narrowly, at a specific university, as was the case with Gross et al. (2018).
The spatial boundary of the target population may be as narrow as is necessary provided that the
units of analysis within those boundaries are representative of the population of interest and are not
unique as compared to the population of interest, and there are sufficient members of the target
population to ensure a suitably sized sample frame can be developed.
SAMPLING FRAME
The sampling frame is an operationalized representation of the target population and is the group of
units from which the sample is recruited. It is the precise group of units – often individuals – that
will be solicited for their participation in the study. Sampling frames may be organized using tele-
phone numbers, names of persons, physical addresses, email addresses, social media groups, organi-
zational lists, or geographical units (Kölln et al., 2019). The nature of the sampling frame is directly
related to the sampling method, as the sampling frame is the operationalized structure through which
the sample is recruited (Prandner & Weichbold, 2019).
Although not desired, it is possible that the sampling frame may extend beyond the boundaries of
the target population and the population of interest, as units that potentially do not meet the inclu-
sion/exclusion criteria may errantly be included. For example, if a human resources department
sends an email participation request to all line employees, it is possible for non-line employees to be
included in the list (e.g., recent promotions, furloughed employees). Such differences between the
sampling frame and the target population (and population of interest) may account for outliers
within the data (Tabachnick & Fidell, 2018). Because the characteristics of natural groups tend to ex-
hibit normal distributions, data points outside of the normal distribution may be indicative of a par-
ticipant who is distinctly different than the natural group. To mitigate the issue of non-qualified units
345
Describing Populations and Samples
of analysis participating in the study, one must clearly define the sampling frame from which the
sample is selected.
A critical characteristic of the sampling frame is that it must include a sufficient number of units that
will support the collection of a complete data set. If the sampling frame is not large enough, the re-
searcher runs the risk of not obtaining sufficient data, resulting in imprecise measures of effect size
for quantitative studies (Hackshaw, 2008) or not reaching data saturation for qualitative studies
(Fusch & Ness, 2015). The size of the sampling frame is a function of the minimum a priori sample
size and the anticipated participation rate.
𝑁𝑁𝑚𝑚𝑚𝑚𝑚𝑚
𝑁𝑁𝑠𝑠𝑠𝑠 =
𝑟𝑟𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝
Where,
N sf - minimum sampling frame size
N min - minimum a priori sample size
rpart - participation rate
The first parameter to determine is the minimum sample size, the description and calculation of
which are detailed in the next section. A common way to determine the minimum sample size for a
quantitative study is using an a priori power analysis based upon the study parameters and the antici-
pated statistical test. The second parameter is the estimated participation rate. The most accurate way
to determine the participation rate is to examine the literature of similar studies that use similar sam-
pling methods as what is proposed. Participation rates vary greatly with general internet advertising
yielding potentially very low rates of less than 1% (Casteel, 2016) to recruitment amongst affiliated
organizations occasionally yielding very strong participation rates, such as 79% rates experienced in
hospital-based research (Bochud et al., 2017). If the literature reports a wide range of participation
rates and one is not able to determine the circumstances that are most aligned with the research, a
recommendation is to underestimate the participation rate and to use the lowest reported rate. In the
absence of any reported participation rates, based on our personal research experience, a guideline
for general recruitment methods is to expect a 5% to 10% participation rate from the sampling
frame.
By way of example, a calculation to determine the minimum sampling frame size requirements for a
study examining the extent to which household turnover rates differ between slum sites. The parame-
ter of minimum sample size, as determined by a priori power analysis and extended to accommodate
outliers and the potential for non-parametric testing, is 148 households. Using an expected participa-
tion rate of 5%, the sampling frame size calculation becomes:
Sampling frame size = 148/.05 = 2,960 households
The calculation means that within the fully described boundaries of the sampling frame, there should
be a minimum of 2,960 eligible potential households meeting the criteria of being located within a
designated slum site, the residents meet the criteria outlined within the description of the target pop-
ulation, and there is a means by which the researcher may contact the potential participant, to include
door-to-door solicitation. There is a great deal to consider when developing and describing the sam-
pling frame. However, the benefit is the time and effort invested into examining and considering all
the elements of sampling frame development will inevitably be rewarded with a well-defined data set
that is appropriate and complete.
Based upon these considerations, the description of the sampling frame should include the opera-
tionalized boundaries of the group, incorporating how the prospective participants are listed, and a
description of how the sampling frame aligns with the sampling method. One must clearly describe
the size of the sampling frame and demonstrate that the sample recruited from the sampling frame
will meet the study data collection requirements.
346
Casteel & Bridier
SAMPLING METHODS
The approach one uses to obtain the members of the sample is the sampling method. The sample is
recruited from the sampling frame in one of two general ways: probability sampling and nonproba-
bility sampling. Common methods used to obtain probability and nonprobability samples are out-
lined below.
P ROBABILITY S AMPLING
Probability sampling describes methods used to develop the sample based upon known probabilities,
allowing one to make inferences about the population of interest following analysis (Sudman, 1976).
Probability sampling is used primarily for quantitative research. Because the sample is developed at
random, the nature of the sample is that it has a probabilistic representation of the population of
interest in which each member of the population has an equal chance of selection (Etikan & Bala,
2017). Below are some of the predominant probability sampling methods used in dissertation re-
search.
347
Describing Populations and Samples
Practically speaking, selection of random participants is accomplished via random number generator.
Using a random number generator, each name is associated with a value. The values and names are
then sorted using the values and the first n names on the sorted list become the random sample.
Systematic sampling
Another method to probabilistically select the sample is systematic sampling. Using this method, one
determines the sample size, n, and the sampling frame, N. Using an ordered list, such as names in a
directory or email addresses in a listserv, one divides the sampling frame into groups of k partici-
pants, where k = N/n. Selecting a random number between 1 and k identifies the first selection of
the sample, then every kth unit on the list becomes a part of the sample. For example, if N = 200
and n = 20, then k = 200/20 = 10, meaning that there are 20 groups of k = 10 members each. Se-
lecting a random integer between 1 and k, where k = 10, may yield a value of 7. The 7th name on the
list becomes the first member of the sample. The second selection is 7 + k, or number 17 on the list.
One continues down the list from 7, 17, 27, 37, …, 197, in which the sample has n = 20 members.
Systematic sampling is also employed in group selection for experimental research. For example, to
form four separate experimental condition groups, the members are listed, and then every fourth
member goes to the same group.
Stratified sampling
When the sampling frame has two or more groups of interests, or strata, then stratified sampling as-
sists in ensuring that each group of interest is proportionally represented (Sudman, 1976). One first
determines the proportion of each group of interest within the sampling frame. For example, at an
undergraduate university, first-year students may represent 33% of the undergraduate student popu-
lation, sophomores 28% of the undergraduate population, juniors may be 22%, and seniors may rep-
resent 17% of the population. Because the first-year students outnumber the graduating seniors
nearly 2:1, it is appropriate that any sample of undergraduate students include nearly twice as many
first-year students as seniors. This may occur if one used simple random sampling; however, the re-
searcher can ensure the proportions are met using stratified sampling.
In stratified sampling the proportions are determined, and then simple random sampling occurs
within each stratum until the proportionate sub-sample size is reached. The simple random sampling
within each stratum allows for probabilistic representation of other (secondary) characteristics that
may not be of interest to the study, while the stratified sampling ensures a probabilistic representa-
tion for undergraduate class, which is a characteristic of interest for the study. For a study seeking a
sample of 200 undergraduate students and using the undergraduate proportions stated above, one
multiplies the total sample size, N, by the proportion to determine the subsample size. For first-year
students, n1 = .33(200) = 66.
Cluster sampling
When a sampling frame is large and diverse, it may at times be economically advantageous to admin-
ister cluster sampling. Cluster sampling occurs when one divides the sampling frame into natural clus-
ters, which are most often geographically bounded areas. Instead of attempting to conduct a simple
random sample of a large, spread out city like Phoenix, Arizona, one might instead divide the city
into clusters, such as by postal zip code. If one selects a random set of postal zip codes to study, one
will be able to use the data within each cluster (zip code) to probabilistically infer information about
the entire city of Phoenix. Phoenix has 47 postal zip codes that cover the city’s 1.45 million inhabit-
ants. One could select eight of the postal zip codes at random. The households within each of the
eight selected postal zip codes are part of the sample. One should note that cluster sampling is often
used to collect large volumes of data and is often used in marketing research, as well as community-
based research.
348
Casteel & Bridier
N ON -P ROBABILITY S AMPLING
Some populations can be hidden from plain sight and do not allow for probability or random sam-
pling. Instead, non-probability sampling is more suited to addressing these populations (Bacher et al.,
2019). Such groups may include ethnic groups, sexual minorities, stigmatized populations, substance
abusers, mobile populations, and others that do not readily identify themselves or that do not present
themselves as distinct or organized populations. To investigate these populations, non-probability
sampling methods are more practical for finding and collecting data to complete research. The bene-
fit of these sampling methods is that they afford access to the necessary groups of people; but this
access is not without shortcomings, as the results of studies that utilize non-probability sampling
methods are not generalizable to the broader population of interest. For example, those who choose
to participate may differ in important ways from those who choose not to participate, providing an
imbalanced or biased view of the population of interest (Kitchenham & Pfleeger, 2002).
Convenience sampling
Convenience sampling is characterized by selecting participants based upon their proximity to the re-
searcher in which the researcher recruits from an opportune sampling frame. An issue arising from
the use of convenience sampling is that the sample rarely represents the population of interest, as it
does not offer the randomness and diversity that exists within the population of interest. Conven-
ience samples often are from the same geographic region, share similar socioeconomic characteris-
tics, and regularly have similar racial or ethnic backgrounds (Emerson, 2015). As such, the results and
conclusions derived from the analysis of a convenience sample may not be generalizable. However,
there are some arguments for limited generalizability based upon the concept of proximal similarity in
which the results are generalizable to a broader audience to the extent that population exhibits similar
characteristics to the sample (Campbell, 1986). A common example of convenience sampling is the
undergraduate class that provides data for their professor. Although such a population makes the
data collection convenient to the researcher, it is not always beneficial to the scientific community.
Meta-analyses examining the differences between college students and nonstudents in research reveal
some stark contrasts in the effects that are reported, both in magnitude and direction, suggesting that
college students are not representative of the adult population as a whole (Peterson, 2001), demon-
strating an obvious weakness of convenience sampling.
Volunteer sampling
Volunteer sampling seeks out participants for a study based upon the participant’s self-selection to
provide data. Unlike convenience sampling in which the localized sampling frame is proactively re-
cruited by the researcher for participation, volunteer sampling occurs when the individual comes
across the opportunity to participate in the research and opts into the study (Fricker, 2016). The char-
acteristic of opting-in to a study is unique to volunteer sampling, as the participant chooses them-
selves to provide data, often based upon an open advertisement or broad-based solicitation for par-
ticipants. All other sampling methods select the participant to participate, and then the participant
can opt-out of the study either by non-response to the participation offer or by non-participation in
part or in whole once included in the study. Volunteer sampling is associated with self-selection bias,
which is a threat to the external validity of the study.
Snowball sampling
Also known as chain sampling, chain-referral sampling, or network sampling, snowball sampling de-
scribes the non-probability sampling method of study participants recruiting future subjects from
amongst those within their sphere of influence (Sharma, 2017). Snowball sampling is effective at
reaching populations that are hard-to-reach or hard-to-ask, allowing members of the hidden popula-
tion to conduct the recruiting on the researcher’s behalf. Snowball sampling is accomplished by the
identification of one or more individuals that represent the population of interest. These individuals
349
Describing Populations and Samples
are then asked to seek participation from others that are like them. The snowball (or chain) continues
as the second-generation participant recruits a third tier of participants meeting the criteria, and so
forth until data collection is complete. There are, however, challenges with snowball sampling. Indi-
viduals who are a part of a hidden population may not be comfortable openly discussing the topic
they are keeping hidden. There is risk to the individual should the topic of study be associated with
their identity, such that the individual may not participate. Researchers may find it difficult to gather
data from individuals who have been referred due to a lack of relationship with or lack of relatability
to the potential participant (Waters, 2015). As with other non-probability samples, the results ob-
tained from snowball sampling are not generalizable to the population of interest due to threats to
external validity. When considering snowball sampling, one must examine the reason for its use and
weigh the risks and challenges that one should expect with its implementation.
Purposive sampling
Also known as judgement sampling, purposive sampling is the intentional selection of a participant
because of the characteristics and qualities the individual possesses (Etikan et al., 2016). The criteria
for selecting the individual may vary, including seeking a specific narrative to explore, a common ex-
perience with a phenomenon, membership in a culture, or being in position to assist in developing a
theory (Creswell, 2013). The common denominator is that each individual is selected because that in-
dividual is known to have a specific quality that is of interest to the investigator. Due to the nature of
the sampling method, purposive sampling is most often seen within qualitative research designs (Pat-
ton, 2015). It bears repeating that purposive sampling methods do not contribute to generalizing re-
sults to the population of interest.
SAMPLE
The sample is the set of units selected to represent the population of interest (Gravetter & Wallnau,
2017). The data provided about the sample will be analyzed and the results inferred (quantitative) or
transferred (qualitative) to the population of interest. The sample should be representative of that
population of interest, a requirement addressed by prescribing the correct sampling frame and by us-
ing an appropriate sampling method. When selecting a sample, there are two primary considerations:
how many units must be in the sample (sample size) and how will these units be selected (sampling
methods). Figure 1 depicts how a researcher identifies a sample from the population of interest and
the target population within the sampling frame.
350
Casteel & Bridier
Note. In the two center ellipses, the target population is depicted by right to left diagonal lines and the sampling
frame is depicted by left to right diagonal lines. The sampling frame intersects the target population. The sam-
ple and sampling frame described extends outside of the target population and population of interest as occa-
sionally the sampling frame may include individuals not qualified for the study.
Figure 1. The relationship between populations within research
The minimum sample size was calculated using a priori power analysis for an in-
dependent-measures t-test (two tails). Based upon an estimated medium effect (d
= .50), level of significance of .05, and a minimum power of .80 with evenly
sized groups (N2/N1 = 1), the minimum sample size is 128 individual licensed
practitioners, with a minimum of 64 in each group.
The minimum sample size calculation states that if you have exactly the predetermined number of
participants (e.g., N = 128), and if those participants were split equally into two groups (n1 = 64, n2 =
64), and if the effect size is exactly as predicted (e.g., d = .50), then the statistical power of the com-
parison is .80. However, the likelihood of this occurring is very low, not because of poor planning –
the calculations are correct – but because real recruiting and data collection, and real effect sizes
rarely work out so perfectly. A good rule of thumb is to seek a minimum sample size that is at least
351
Describing Populations and Samples
15% greater than the calculated minimum sample size from the a priori power analysis. Using the
previous example, this means the sample size of 128 becomes 148 (128 * 1.15 = 147.2, round up to
the nearest whole number). Recruiting to this sample size accommodates two primary considerations:
1. Data often include outliers, which may be removed to conform to parametric analysis as-
sumptions, thus reducing the effective sample size. Outlier data may also be transformed to
conform to the data as a preferred alternative to removing data (Tabachnick & Fidell, 2018).
If one were to have collected the exact-calculated minimum sample size and outliers occur
and are removed, one begins the analysis with a statistical power deficit due to a smaller-
than-anticipated sample size.
2. The data may be non-parametric, requiring the use of a non-parametric analysis. In the case
where the data violate one or more of the assumptions for the selected parametric test, the
use of the non-parametric alternative may be required. Because non-parametric tests are
based upon rank ordering rather than distributions and the associated probability functions,
a larger sample size is required to achieve the same statistical power as its parametric coun-
terpart. A sample size that is 15% greater than the calculation for the parametric a priori
power analysis provides the equivalent statistical power for the associated non-parametric
test (Lehmann, 1973). For example, the minimum sample size based upon the power analysis
for an independent measures t-test (two tails) may be N = 128 with a power of .80, and a
Mann-Whitney U test requires N = 148 to achieve the same power of .80 (min relative effi-
ciency), provided all other parameters are equal (e.g., level of significance, effect size).
The second method of determining sample size is mostly used within quantitative descriptive designs
in which one is interested in describing the characteristics of the population of interest and when
representing all members of that group are a priority. This second method uses confidence level cal-
culations, which provides the minimum sample size one must recruit to meet the desired statistical
constraints; namely, the sample represents the characteristics of interest for the entire population of
interest.
A confidence level describes the extent to which the selected sample probabilistically represents the
population of interest. Naturally, the higher the confidence level, the higher the probability the sam-
ple is representative of that population. Of course, the limitation on sampling at a higher confidence
level is that it requires a larger sample size, which imposes restrictions upon one’s resources (e.g., fi-
nances, time, people). Regardless of the approach – unless sampling the entire population of interest,
such as the U.S. Census – there will be errors due to the probabilistic nature of the sampling. The er-
rors associated with the variation between samples due to chance are called sampling error. The smaller
the sampling error, the less variation between random samples from the same population of interest.
However, to achieve smaller sampling errors, one must increase the sample size from the population
of interest (Cohen, 2013). The margin of error is the maximum expected error based upon the level
of confidence.
To calculate the sample size based upon confidence levels, one may use the following equation:
𝑧𝑧 2 × 𝑝𝑝̂ (1 − 𝑝𝑝̂ )
𝑁𝑁 =
ε2
where,
Confidence level desired (z) – The probability, as expressed by the z-score, the selected
sample represents the population of interest
Margin of error (𝜀𝜀) – The variation between samples selected from the same population of
interest, a percentage expressed as a probability (e.g., 𝜀𝜀 = .05).
Population of interest size (N) – The measured or estimated size of the population of in-
terest
352
Casteel & Bridier
353
Describing Populations and Samples
needed before data saturation is reached being the amount of data collected and how rich with infor-
mation it is.
Although this rationale may seem overly simplistic, it is nonetheless the case. Research that incorpo-
rates lengthy individual open-ended interviews as a principal source of data will achieve data satura-
tion with fewer participants than a study solely relying upon written responses from its participants.
Because interviews – as compared to written responses or other survey data sources – enable the re-
searcher to ask specific questions to guide the discussion and provide opportunities for real-time ex-
ploration of participant responses, interview data will tend to be of much richer quality than written
sources. Of course, the cost of obtaining rich sources of data, such as individual interviews and fo-
cus groups, is the time expenditure of the researcher and participant to develop a quality set of data.
An additional consideration is the skillset and experience of the interviewer to conduct the interview,
notice opportunities to explore openings in the participant’s descriptions, ask probing questions to
investigate the nature of the phenomenon, pick up on participant nonverbal cues, and to measure the
development and repetition of themes real-time to guide the discussion and recognize natural con-
clusion points (McGrath et al., 2019; Tracy, 2019). The depth of information required by the research
problem and research design, the source(s) of data, and the skills of the researcher are all important
considerations, amongst others, in determining the number of participants one would anticipate
needing to reach the goal of data saturation.
The scope of the research problem also must be considered in determining the number of partici-
pants one would likely need to develop quality themes. Topics that are narrow in scope, seek rela-
tively superficial levels of examination, and have low inference, such as in descriptive designs, require
only a moderate amount of data (Sandelowski, 2000). Therefore, the themes are likely to be devel-
oped and reach saturation after modest participation. Alternatively, examination of deep, rich, and
meaningful phenomena requires a greater amount of data to understand the core experience. Such
examination is a balance between sample size and the amount of data collected via interview. The
study of cultures through ethnography requires larger sample sizes with moderate interview lengths
to achieve the necessary depth and breadth of cultural understanding (Bernard, 2013). Similarly,
grounded theory research typically involves a moderate number of participants, as the purpose is to
develop a well-saturated theory. (Creswell, 2013). Phenomenological studies tend to focus upon
depth of understanding of the lived experience, requiring significant contact time between the re-
searcher and participant (Marshall et al., 2013). The natural focus tends to limit the number of partic-
ipants involved in phenomenological studies while emphasizing the number, length, and depth of in-
terviews. Narrative study, by nature, involves few participants, but engages in deep examination of
the story to be shared (Riessman & Quinney, 2005). However, larger studies examining broader expe-
riences are certainly evident in the literature (e.g., N = 18; Gray et al., 2002). Case studies are a some-
what unique in that the unit of analysis is the case, as opposed to the phenomenon. As such, sample
selection for case studies depends upon the nature of the case. In some circumstances, the case is
limited to the experiences of a single person (Yin, 2014), while other cases may be centered around
events, groups, communities, programs, decision points, implementations of new processes, or other
experiential inflection points. The number of participants required for each case is clearly dependent
upon the nature of the case itself.
Because these guidelines are developed with the novice researcher in mind, the skills of the re-
searcher will be considered as beginning level. Therefore, the research design and the source of data
will be the primary considerations for describing recommendations for sample recruitment in qualita-
tive research.
Descriptive . Kim et al. (2017) pointed out that a majority of descriptive design studies (60%) in-
cluded the data of 20 or fewer participants, with 44% of the surveyed studies having between 11 and
20 participants. The synthesis of the literature survey and the nature of the design suggest that a
range of 11 to 20 participants is an appropriate starting point for recruitment in descriptive research
studies that use interviews or focus groups as primary sources of data. This recommendation also
354
Casteel & Bridier
aligns with the writings of Lincoln and Guba (1985), who suggest 12-20 participants for interview
studies. Although interview lengths will depend upon the research problem, the superficial nature of
the design suggests that 45-75 minutes of dialogue should provide adequate data for novice research-
ers.
Case study . With a predominance of cases being single individuals (Yin, 2014), it is feasible to re-
cruit one person as the sample. However, small sample sizes are often too small to support develop-
ment of claims, provide informational redundancy, or acquire theoretical saturation (Sandelowski,
1995). Creswell (2013) recommends selecting “cases that show different perspectives on the problem,
process or event” (p. 100), suggesting multiple participants must be included, while also recommend-
ing multiple cases to be used to investigate the problem. Marshall et al. (2013) go on to recommend
that saturation often occurs at 30 interviews. Because case study highly recommends the use of mul-
tiple additional sources to corroborate the data, reliance upon the interview as the sole source of data
is not required; therefore, the number of interview participants can be lower than recommended by
Marshall et al. As an accommodation between the need for extensive interview data and the use of
additional sources of data to seek saturation, including 12-15 participants are recommended to pro-
vide multiple perspectives of the case phenomenon while using additional sources of data to support
the findings. Interview lengths may be similar to those of descriptive research, ranging from 45-75
minutes.
Phenomenolog y . The guidelines for participant recruitment for phenomenological studies vary
across the literature, with Marshall et al. (2013) reporting ranges from 6-10 participants as a recom-
mendation. It is also shared in the literature that samples sizes of one for phenomenological study
are not rare (Sim et al., 2018). Smith and Osborn (2003) note that five or six participants have been
recommended as an appropriate size for student studies using interpretive phenomenology and go
on to offer their own recommendation of three participants for first-time researchers. The consensus
amongst the phenomenological authors suggests a recommendation of 3-10 participants recruited as
part of a phenomenological design for novice researchers with consideration given to having at least
six complete data sets (participants). Contact time with participants will be extensive within phenom-
enological designs. Expect for interviews to be a minimum of 90-120 minutes; longer, if fewer par-
ticipants are interviewed. Conducting multiple interviews with the same participant is a regular char-
acteristic of the design.
Ethnog raphy. The recommendations of Bernard (2011) are for 30-60 participants in ethnographic
studies to provide the broad range of views required to achieve cultural understanding. However,
Bernard (2013, p. 175) also intimates that “10-20 knowledgeable people are enough to uncover and
understand the core categories in any well-defined cultural domain or study of lived experience.”
Novice researchers would be well-suited to consider between 20-30 participants when examining a
core principle; more if the construct of examination is broader. Interviews may be between 60-120
minutes with follow-up interviews with the same participants for clarification and to fill in gaps in the
social and cultural patterns.
Grounded theory . Because grounded theory seeks to develop a unified theoretical explanation of a
process (Corbin & Strauss, 2007), a larger number of participants is requisite to develop and support
the emerging theory. A minimum recommendation for sample sizes of 20-30 participants is put forth
by Creswell (2013), while Charmaz (2014) advocates for larger samples. Grounded theory uses inter-
views with the study participants to initially develop the data from which the theory is created, and
then requires regular and repeated contacts with the same participants to compare findings, fill gaps,
and add to the framework. Initial interviews may last 60-120 minutes with follow-up interviews rang-
ing in time from 15-60 minutes.
Narrative . Narrative research seeks to tell the story of an individual or small group so as to provide
a sequence to events and explain the consequences and connections of those events (Riessman &
Quinney, 2005). Creswell (2013) recommends two to three participants for narrative research, as the
355
Describing Populations and Samples
interviews tend to be longer in duration. Groleau et al. (2006) report interview lengths averaging 120
minutes for narrative study.
It should be noted that the recommendations from the literature are the participation requirements
to obtain sufficient data to complete a thorough data analysis. Some colleges and universities may re-
quire their graduate students to obtain a greater minimum sample size than is recommended above,
and for good reason. Learning programs, such as masters and doctoral curricula, seek to ensure that
the student learns the appropriate interview skills, which is something that does not happen without
practice. Therefore, the inclusion of additional participants affords the novice researcher the oppor-
tunity to learn interviewing skills as the research progresses. Additionally, these minimum sample
sizes are put in place to ensure that the student has sufficient data to conduct an analysis. It is not un-
common for students to come back from the field with interviews that are too short or have exten-
sive transcripts of the interviewer speaking and minimal contribution from the participant. Although
the use of appropriate interview and re-interview methods should alleviate the need for these mini-
mums, and field tests assist in developing interview skills, the minimums are set in place as a stop gap
to mitigate the risk of insufficient data for analysis. Table 1 provides an overview of qualitative de-
signs with recommended sample sizes and sources for each.
Table 1. Recommended a priori sample size for qualitative research designs
356
Casteel & Bridier
357
Describing Populations and Samples
CONCLUSION
Social science research is the study of people, organizations, groups, or networks of people, and to
effectively conduct such studies, one must effectively and thoroughly define and describe what is be-
ing examined. The descriptions of the population and sample, as well as the defining elements, such
as the unit of analysis and unit of observation, serve as a foundation from which the rest of the
study is designed. Using an intentional approach of defining and describing each element within the
dissertation aids doctoral students in clearly describing the subject of research, as well as demon-
strates strong understanding of how population and samples apply to general research design.
REFERENCES
Ackerman, B., Schmid, I., Rudolph, K. E., Seamans, M. J., Susukida, R., Mojtabai, R., & Stuart, E. A. (2019).
Implementing statistical methods for generalizing randomized trial findings to a target population. Addictive
Behaviors, 94, 124-132. https://doi.org/10.1016/j.addbeh.2018.10.033
Bacher, J., Lemcke, J., Schmich, P., & Quatember, A. (2019). Probability and nonprobability sampling: Repre-
sentative surveys of hard-to-reach and hard-to-ask populations. Current surveys between the poles of the-
ory and practice. Survey Methods: Insights from the Field. https://doi.org/10.13094/SMIF-2019-00018
Bakibinga, P., Kabaria, C., Kyobutungi, C., Manyara, A., Mbaya, N., Mohammed, S., Njeri, A., Azam, I., Iqbal,
R., Mazaffar, S., Rizvi, N., Rizvi, T., Rehman, H. U., Shifat Ahmed, S., Alam, O., Khan, A. Z., Rahman, O.,
Yusuf, R., Odubanjo, D., … Yeboah, G. (2019). A protocol for a multi-site, spatially-referenced household
survey in slum settings: Methods for access, sampling frame construction, sampling, and field data collec-
tion. BMC Medical Research Methodology, 19(1), 109. https://doi.org/10.1186/s12874-019-0732-x
Bandura, A. (1977). Social learning theory. Prentice-Hall.
Baxter, P., & Jack, S. (2008). Qualitative case study methodology: Study design and implementation for novice
researchers. The Qualitative Report, 13(4), 544-559. https://nsuworks.nova.edu/tqr/vol13/iss4/2
Bernard, H. R. (2011). Research methods in anthropology: Qualitative and quantitative approaches (5th ed.). Rowman &
Littlefield.
Bernard, H. R. (2013). Social research methods: Qualitative and quantitative approaches (2nd ed.). Sage.
Blaikie, N. (2018). Confounding issues related to determining sample size in qualitative research. International
Journal of Social Research Methodology, 21(5), 635-641. https://doi.org/10.1080/13645579.2018.1454644
Bochud, M., Currat, C., Chapatte, L., Roth, C., & Mooser, V. (2017). High participation rate among 25,721 pa-
tients with broad age range in a hospital-based research project involving whole-genome sequencing-the
Lausanne Institutional Biobank. Swiss Medical Weekly, 147. https://doi.org/10.4414/smw.2017.14528
358
Casteel & Bridier
Campbell, D. T. (1986). Relabeling internal and external validity for applied social scientists. New Directions for
Program Evaluation, 1986(31), 67-77. https://doi.org/10.1002/ev.1434
Casteel, B. A. III (2016). Relationships between learners’ personality traits and transactional distance within an e-learning envi-
ronment (Doctoral dissertation, Grand Canyon University). ProQuest Dissertations and Theses (Publication
No. 10159244).
Charmaz, K. (2014). Constructing grounded theory (2nd ed.). Sage.
Coase, R. H. (1937). The nature of the firm. Economica, 4, 386-405. https://doi.org/10.2307/2626876
Cohen, J. (2013). Statistical power analysis for the behavioral sciences (2nd ed.). Psychology Press.
https://doi.org/10.4324/9780203771587
Colorafi, K. J., & Evans, B. (2016). Qualitative descriptive methods in health science research. Health Environ-
ments Research & Design Journal, 9(4), 16-25. https://doi.org/10.1177/1937586715614171
Corbin, J., & Strauss, A. (2007). Basics of qualitative research: Techniques and procedures for developing grounded theory
(3rd ed.). Sage. https://doi.org/10.4135/9781452230153
Creswell, J. A. (2013). Qualitative inquiry and research design: Choosing among five approaches (3rd ed.). Sage.
Deci, E. L., & Ryan, R. M. (2008). Self-determination theory: A macrotheory of human motivation, develop-
ment, and health. Canadian Psychology/Psychologie Canadienne, 49(3), 182-185.
https://doi.org/10.1037/a0012801
Dieumegard, G., Nogry, S., Ollagnier-Beldame, M., & Perrin, N. (2019). Lived experience as a unit of analysis
for the study of learning. Learning, Culture and Social Interaction, 100345.
https://doi.org/10.1016/j.lcsi.2019.100345
Emerson, R. W. (2015). Convenience sampling, random sampling, and snowball sampling: How does sampling
affect the validity of research? Journal of Visual Impairment & Blindness, 109(2), 164-168.
https://doi.org/10.1177/0145482X1510900215
Erickson, F. (1984). What makes school ethnography ‘ethnographic’? Anthropology & Education Quarterly, 15(1),
51-66. https://doi.org/10.1525/aeq.1984.15.1.05x1472p
Etikan, I., & Bala, K. (2017). Sampling and sampling methods. Biometrics & Biostatistics International Journal, 5(6),
215-217. https://doi.org/10.15406/bbij.2017.05.00149
Etikan, I., Musa, S. A., & Alkassim, R. S. (2016). Comparison of convenience sampling and purposive sampling.
American Journal of Theoretical and Applied Statistics, 5(1), 1-4. https://doi.org/10.11648/j.ajtas.20160501.11
Fricker, R. D. (2016). Sampling methods for online surveys. In N. G. Fielding, R. M. Lee, & G. Blank (Eds.), The
SAGE handbook of online research methods (2nd ed., pp. 162-183). Sage.
https://doi.org/10.4135/9781473957992.n10
Fusch, P., & Ness, L. (2015). Are we there yet? Data saturation in qualitative research. The Qualitative Report,
20(9), 1408-1416. https://nsuworks.nova.edu/tqr/vol20/iss9/3
Giorgi, A. (2006). The descriptive phenomenological method in psychology: A modified Husserlian approach. Duquesne Uni-
versity Press.
Gravetter, F. J., & Wallnau, L. B. (2017). Statistics for the behavioral sciences (10th ed.). Wadsworth.
Gray, R. E., Fitch, M. I., Fergus, K. D., Mykhalovskiy, E., & Church, K. (2002). Hegemonic masculinity and ex-
perience of prostate cancer: A narrative approach. Journal of Aging and Identity, 7(1), 43-62.
https://doi.org/10.1023/A:1014310532734
Groleau, D., Young, A., & Kirmayer, L. J. (2006). The McGill Illness Narrative Interview (MINI): An interview
schedule to elicit meanings and modes of reasoning related to illness experience. Transcultural Psychiatry,
43(4), 671-691. https://doi.org/10.1177/1363461506070796
Gross, M., Moore, Z. E., Gardner, F. L., Wolanin, A. T., Pess, R., & Marks, D. R. (2018). An empirical examina-
tion comparing the mindfulness-acceptance-commitment approach and psychological skills training for the
359
Describing Populations and Samples
mental health and sport performance of female student athletes. International Journal of Sport and Exercise
Psychology, 16(4), 431-451. https://doi.org/10.1080/1612197X.2016.1250802
Grünbaum, N. N. (2007). Identification of ambiguity in the case study research typology: What is a unit of
analysis? Qualitative Market Research: An International Journal, 10(1), 78-97.
https://doi.org/10.1108/13522750710720413
Hackshaw, A. (2008). Small studies: Strengths and limitations. European Respiratory Journal, 32(5), 1141-1143.
https://doi.org/10.1183/09031936.00136408
Hammersley, P., & Mairs, H. (2004). Sampling methods. Nurse Researcher, 12(1), 4-6.
https://doi.org/10.7748/nr2004.10.12.1.4.c5926
Kalleberg, A. L., Marsden, P. V., Aldrich, H. E., & Cassell, J. W. (1990). Comparing organizational sampling
frames. Administrative Science Quarterly, 35(4), 658–688. https://doi.org/10.2307/2393513
Kenny, D. A., Kashy, D. A., & Cook, W. L. (2006). Dyadic data analysis. The Guilford Press.
Khan, S. N. (2014). Qualitative research method: Grounded theory. International Journal of Business and Manage-
ment, 9(11), 224-233. https://doi.org/10.5539/ijbm.v9n11p224
Kim, H., Sefcik, J. S., & Bradway, C. (2017). Characteristics of qualitative descriptive studies: A systematic re-
view. Research in Nursing & Health, 40(1), 23-42. https://doi.org/10.1002/nur.21768
Kitchenham, B., & Pfleeger, S. L. (2002). Principles of survey research: Part 5: Populations and samples. ACM
SIGSOFT Software Engineering Notes, 27(5), 17-20. https://doi.org/10.1145/571681.571686
Kölln, A. K., Ongena, Y. P., & Aarts, K. (2019). The effects of sampling frame designs on nonresponse and
coverage error: Evidence from the Netherlands. Journal of Survey Statistics and Methodology, 7(3), 422-439.
https://doi.org/10.1093/jssam/smy016
Kumar, S. (2018). Understanding different issues of unit of analysis in a business research. Journal of General
Management Research, 5(2), 70-82. https://www.bloomsbury.com/in/journal-of-general-management-re-
search-9789389000207/
Lehmann, E. L. (1973). Non-parametrics: Statistical methods based on ranks. Holden-Day.
Lewis, L. (2020). Finding the stories: A novice qualitative researcher learns to analyse narrative inquiry data.
Nurse researcher, 28(1). https://doi.org/10.7748/nr.2018.e1578
Lincoln, Y. S., & Guba, E. G. (1985). Naturalistic inquiry. Sage. https://doi.org/10.1016/0147-1767(85)90062-8
Magilvy, J. K., & Thomas, E. (2009). A first qualitative project: Qualitative descriptive design for novice re-
searchers. Journal of Specialists in Pediatric Nursing, 14(4), 298-300. https://doi.org/10.1111/j.1744-
6155.2009.00212.x
Marks, A., Wilkes, L., Blythe, S., & Griffiths, R. (2017). A novice researcher’s reflection on recruiting partici-
pants for qualitative research. Nurse Researcher, 25(2), 34. https://doi.org/10.7748/nr.2017.e1510
Marshall, B., Cardon, P., Poddar, A., & Fontenot, R. (2013). Does sample size matter in qualitative research? A
review of qualitative interviews in IS research. Journal of Computer Information Systems, 54(1), 11-22.
https://doi.org/10.1080/08874417.2013.11645667
McCrae, R., & Costa, T. (2003). Personality in adulthood: A Five-Factor Theory perspective (2nd ed.). Guilford.
https://doi.org/10.4324/9780203428412
McGrath, C., Palmgren, P. J., & Liljedahl, M. (2019). Twelve tips for conducting qualitative research interviews.
Medical Teacher, 41(9),1002-1006. https://doi.org/10.1080/0142159X.2018.1497149
Meyer, J., & Allen, N. (1997). Commitment in the workplace: Theory, research, and application. Sage.
https://doi.org/10.4135/9781452231556
Neuman, W. L. (2013). Social research methods: Qualitative and quantitative approaches. Sage.
Ojala, T., Häkkinen, A., Karppinen, J., Sipilä, K., Suutama, T., & Piirainen, A. (2014). Chronic pain affects the
whole person – A phenomenological study. Disability and Rehabilitation, 37(4), 363-371.
https://doi.org/10.3109/09638288.2014.923522
360
Casteel & Bridier
Patton, M. Q. (2015). Qualitative research and evaluation methods (4th ed.). Sage.
Peterson, R. A. (2001). On the use of college students in social science research: Insights from a second-order
meta-analysis. Journal of Consumer Research, 28(3), 450-461. https://doi.org/10.1086/323732
Porzsolt, F., Wiedemann, F., Becker, S. I., & Rhoads, C. J. (2019). Inclusion and exclusion criteria and the prob-
lem of describing homogeneity of study populations in clinical trials. BMJ Evidence-Based Medicine, 24(3),
92-94. https://doi.org/10.1136/bmjebm-2018-111115
Prandner, D., & Weichbold, M. (2019). Building a sampling frame for migrant populations via an onomastic
approach: Lesson learned from the Austrian Immigrant Survey 2016. Survey Methods: Insights from the Field,
1-20. https://doi.org/10.13094/SMIF-2019-00006
Riessman, C. K., & Quinney, L. (2005). Narrative in social work: a critical review. Qualitative Social Work, 4(4),
391-412. https://doi.org/10.1177/1473325005058643
Ritella, G., Rajala, A., & Renshaw, P. (2020). Using chronotope to research the space-time relations of learning
and education: Dimensions of the unit of analysis. Learning, Culture and Social Interaction.
https://doi.org/10.1016/j.lcsi.2020.100381
Salkind, N. J. (Ed.). (2010). Encyclopedia of research design. Sage. https://doi.org/10.4135/9781412961288
Sandelowski, M. (1995). Sample size in qualitative research. Research in Nursing & Health, 18(2), 179-183.
https://doi.org/10.1002/nur.4770180211
Sandelowski, M. (2000). Whatever happened to qualitative description? Research in Nursing & Health, 33. 334-
340. https://doi.org/10.1002/1098-240X(200008)23:4<334::AID-NUR9>3.0.CO;2-G
Sharma, G. (2017). Pros and cons of different sampling techniques. International Journal of Applied Research, 3(7),
749-752. https://www.allresearchjournal.com/archives/2017/vol3issue7/PartK/3-7-69-542.pdf
Sim, J., Saunders, B., Waterfield, J., & Kingstone, T. (2018). Can sample size in qualitative research be deter-
mined a priori? International Journal of Social Research Methodology, 21(5), 619-634.
https://doi.org/10.1080/13645579.2018.1454643
Simui, F., Kasonde-Ngandu, S., Cheyeka, A., & Kakana, F. (2018). Unearthing dilemmas in thesis titles: Lived
experience of a novice researcher in Sub-Saharan Africa. International Journal of Multidisciplinary Research and
Development, 5(4), 99-105. http://www.allsubjectjournal.com/archives/2018/vol5/issue4/5-3-46
Smith, J. A., & Osborn, M. (2003). Interpretative phenomenological analysis. In J. A. Smith (Ed.), Qualitative psy-
chology: A practical guide to research methods (pp. 51-80). Sage.
Soto, C. J., & John, O. P. (2017). The next Big Five Inventory (BFI-2): Developing and assessing a hierarchical
model with 15 facets to enhance bandwidth, fidelity, and predictive power. Journal of Personality and Social
Psychology, 113(1), 117-143. https://doi.org/10.1037/pspp0000096
Sudman, S. (1976). Applied sampling. Academic Press.
Tabachnick, B. G., & Fidell, L. S. (2018). Using multivariate statistics (6th ed.). Pearson.
Tracy, S. J. (2019). Qualitative research methods: Collecting evidence, crafting analysis, communicating impact. John Wiley.
Waters, J. (2015). Snowball sampling: A cautionary tale involving a study of older drug users. International Journal
of Social Research Methodology, 18(4), 367-380. https://doi.org/10.1080/13645579.2014.953316
Wohlfart, O. (2020). “Digging deeper?”: Insights from a novice researcher. International Journal of Qualitative
Methods, 19. https://doi.org/10.1177/1609406920963778
Yin, R. K. (2014). Case study research: Design and methods (5th ed.). Sage.
361
Describing Populations and Samples
AUTHORS
Alex Casteel (https://orcid.org/0000-0002-5352-0523) is an assistant
professor and academic quality peer reviewer at Grand Canyon Univer-
sity, College of Doctoral Studies, in Phoenix, Arizona. In these roles, he
teaches research design, statistics, and quantitative methods, and evaluates
dissertation manuscripts, ensuring those manuscripts meet the standards
of quality and scientific merit as set forth by the University for students
in the EdD, PhD, and DBA programs. His research interests include the
role of psychological constructs within online learning environments,
online pedagogy, student retention within online environments, and im-
plementation of technology-based learning solutions.
362