You are on page 1of 10

1098240x, 2000, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1098-240X(200006)23:3<246::AID-NUR9>3.0.CO;2-H by Cochrane Philippines, Wiley Online Library on [25/01/2023].

See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Research in Nursing & Health, 2000, 23, 246–255

Focus on Research Methods


Combining Qualitative
and Quantitative Sampling,
Data Collection,
and Analysis Techniques
in Mixed-Method Studies
Margarete Sandelowski*

University of North Carolina at Chapel Hill, School of Nursing,


7460 Carrington Hall, Chapel Hill, NC 27599
Received 12 July 1999; accepted 13 January 2000

Abstract: Researchers have increasingly turned to mixed-method techniques to expand the


scope and improve the analytic power of their studies. Yet there is still relatively little direction on
and much confusion about how to combine qualitative and quantitative techniques. These tech-
niques are neither paradigm- nor method-linked; researchers’ orientations to inquiry and their
methodological commitments will influence how they use them. Examples of sampling combina-
tions include criterion sampling from instrument scores, random purposeful sampling, and strat-
ified purposeful sampling. Examples of data collection combinations include the use of instru-
ments for fuller qualitative description, for validation, as guides for purposeful sampling, and as
elicitation devices in interviews. Examples of data analysis combinations include interpretively
linking qualitative and quantitative data sets and the transformation processes of qualitizing and
quantitizing. © 2000 John Wiley & Sons, Inc. Res Nurs & Health 23:246–255, 2000

Keywords: mixed-method studies; combination studies; qualitative and quantitative research;


sampling; data collection; data analysis; triangulation

The idea of mixing qualitative and quantitative how of studies combining qualitative and quanti-
methods has stimulated much interest and debate tative techniques.
(e.g., Greene & Caracelli, 1997a; Sandelowski,
1995; Swanson, 1992; Tashakkori & Teddlie,
1998). Researchers increasingly have used mixed-
method techniques to expand the scope of, and THE WHERE, WHY, AND WHAT
deepen their insights from, their studies. As advo- OF COMBINATIONS
cates of mixed-method research have argued, the
complexity of human phenomena mandates more Combination or mixed-method studies are con-
complex research designs to capture them. De- cretely operationalized at the technique level, or
spite this interest, there is still relatively little di- the shop floor, of research: that is, at the level of
rection on and much confusion about how to ac- sampling, data collection, and data analysis.
complish mixed-method studies. In this paper, I Mixed-method studies are not mixtures of para-
discuss the where, what, why, and especially, the digms of inquiry per se, but rather paradigms are

*Professor.

246 © 2000 John Wiley & Sons, Inc.


1098240x, 2000, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1098-240X(200006)23:3<246::AID-NUR9>3.0.CO;2-H by Cochrane Philippines, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
COMBINING / SANDELOWSKI 247

reflected in what techniques researchers choose to That is, it is uncertain whether a researcher can
combine, and how and why they desire to combine conduct a project framed first, for example, in pos-
them. itivism and then later, in critical theory. Such
worldviews, like all strongly held belief systems,
At the Paradigm Level are not easily exchanged.
As data collection and analysis techniques are
As many scholars (e.g., Guba & Lincoln, 1994; not linked to paradigms (Berman, Ford-Gilboe, &
Heron & Reason, 1997) have variously named and Campbell, 1998; Sandelowski 1995), both the re-
described them, paradigms of inquiry are world- searcher in a positivist viewing position and the re-
views that signal distinctive ontological (view of searcher in a critical theory viewing position may
reality), epistemological (view of knowing and the use interviews and even the very same standard-
relationship between knower and to-be-known), ized measures to answer their questions, but they
methodological (view of mode of inquiry), and ax- will employ these techniques and, more important-
iological (view of what is valuable) positions. In- ly, analytically treat their results differently. In-
deed, paradigms of inquiry are best understood as deed, arguably the key difference between “quali-
viewing positions: ways, and places from which, tative” and “quantitative” (when these words are
to see. Because different paradigms, such as (neo- used to signal different inquiry paradigms or re-
or post-) positivism, constructivism, critical theo- search traditions) researchers is in their attitude to-
ry, and participatory inquiry, entail contradictory ward and treatment of data. Accordingly, although
viewing positions, combinations at the paradigm techniques can be mixed, the resulting mix will re-
level are not true combinations, mergers, or rec- veal the researcher’s viewing position (or, in cases
onciliations of worldviews, but rather the explicit of mixed-up research, the researcher’s futile effort
framing of inquiry in two or more worldviews, to mix the research equivalent of oil and water).
each of which remains distinct from the other. That
is, for example, it is not possible to combine, merge, At the Method Level
or reconcile a view of reality as singular and ob-
jective (positivist) with views of it as multiple and Combinations at the method level can be used to ex-
individually or culturally constructed (construc- pand the scope of a study as researchers seek to cap-
tivist), or as historically contingent (critical theo- ture method-linked dimensions of a target phenom-
ry). A critical theorist will frame the issues around enon (Greene et al., 1989, p. 259). As Wolfer (1993)
hormone replacement for midlife women differ- proposed, different aspects of reality lend them-
ently from a neopositivist.1 That is, they will each selves to different methods of inquiry. For example,
see different things and, therefore, ask different the physiology of hormone replacement therapy for
questions that will, in turn, require the use of dif- women lends itself to study via physiologic mea-
ferent methods and techniques to answer them. sures, while the debate around the benefits and lia-
Such a paradigm combination may be used to elic- bilities of hormone replacement therapy lends itself
it two or more perspectives on hormone therapy, to discourse, semiotic, and other cultural analyses.
or for the purpose of “initiation” (Greene, Cara- In the study I conducted with my colleagues of the
celli, & Graham, 1989, p. 259), to surface para- transition to parenthood of infertile couples (Sande-
doxes and contradictions that surround hormone lowski, Holditch-Davis, & Harris, 1992), we used
therapy. Accordingly, two or more paradigms of grounded theory methodology with naturalistic,
inquiry can be used to frame the same target phe- ethological observation to capture different features
nomenon (in this case, hormone replacement). of this phenomenon, including couples’ under-
Yet, arguably, the positivist and the critical theo- standings of their struggle to become parents and
rist may not really be studying the same phenom- their interaction with their babies. Grounded theo-
enon, because to see a phenomenon in a certain ry and ethological observation share a common nat-
way is to change that phenomenon. As Pearce uralist imperative (that is, not to manipulate, or im-
(1971, p. 2) observed, “a change of world view can pose a priori conceptualizations on, the target
change the world viewed.” Moreover, it is debat- phenomenon) that make them compatible for use in
able whether one researcher can hold two differ- one study. But grounded theory is particularly well-
ent viewing positions, albeit at different times. suited to theorizing human understandings, while
ethological observation is especially well-suited to
1I use terms like “positivist” and “positivism” to quickly com-
theorizing human behavior.
municate complex and admittedly controversial ideas for the Although the various dimensions of phenome-
purposes of this paper, not to stereotype researchers or view- na may be method-linked in that different dimen-
ing positions. sions may be best captured by different methods,
1098240x, 2000, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1098-240X(200006)23:3<246::AID-NUR9>3.0.CO;2-H by Cochrane Philippines, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
248 RESEARCH IN NURSING & HEALTH

methods, like paradigms, are not specifically and quantitative approaches may be used sequen-
linked to techniques. Grounded theory may be tially, concurrently and iteratively, or in a sand-
generated using an array of qualitative and quan- wich pattern.
titative data collection techniques and sources.
Moreover, methods are not uniformly linked to Combining Sampling Strategies
paradigms (Greene & Caracelli, 1997b). Grounded
theory may be conducted in a neopositivist or con- One of the most important features distinguishing
structivist paradigm (Annells, 1996). A neoposi- what is commonly referred to as qualitative from
tivist researcher conducting grounded theory be- quantitative inquiry is the kind of sampling used.
lieves in an external and objectively verifiable While qualitative research typically involves pur-
reality. In contrast, a constructivist conducting poseful sampling to enhance understanding of the
grounded theory believes in multiple, experien- information-rich case (Patton, 1990), quantitative
tially based, and socially constructed realities. For research ideally involves probability sampling to
the neopositivist, concepts emerge or are discov- permit statistical inferences to be made. Although
ered, as if they were there to be found. The act of purposeful sampling is oriented toward the devel-
discovery is separate from that which is discov- opment of idiographic knowledge—from gener-
ered. For the constructivist, concepts are made, alizations from and about individual cases—
fashioned, or invented from data. What construc- probability sampling is oriented toward the devel-
tivists find is what they made. For the construc- opment of nomothetic knowledge, from general-
tivist, all human discovery is creation. izations from samples to populations. Notwith-
standing these key differences, purposeful and
probability sampling techniques can be combined
At the Technique Level usefully.
The technique level of research is the site where Criterion sampling. For example, in design
combinations actually occur and is what is most template 2 shown in Figure 1, in which the use of
often referred to in discussions of mixed-method quantitative techniques precede the use of qualita-
research. Such combinations entail the use of sam- tive techniques, research participants’ scores on
pling, data collection, and data analysis techniques the instruments used to collect data in the quanti-
commonly (although not necessarily) conceived tative portion of the study can be used to initiate a
as qualitative or quantitative. Because techniques criterion sampling strategy. Criterion sampling is
are tied neither to paradigms nor to methods, com- a kind of purposeful sampling of cases on precon-
binations at the technique level permit innovative ceived criteria, such as scores on an instrument.
uses of a range of techniques for a variety of pur- Cases may be chosen because they typify the av-
poses. Three purposes include (a) triangulation, to erage score; this kind of sampling may also be re-
achieve or ensure corroboration of data, or con- ferred to as typical case sampling. Cases may be
vergent validation; (b) complementarity, to clari- chosen because they exemplify extreme scores;
fy, explain, or otherwise more fully elaborate the this kind of sampling may also be called extreme
results of analyses; and (c) development, to guide or deviant case sampling. (The term deviant here
the use of additional sampling, and data collection refers to any departure from a specified norm.)
and analysis techniques (Greene et al., 1989, p. Such cases are highly unusual. Or, cases may be
259). The rest of this paper is devoted to illustrat- chosen because they show a variable intensely, but
ing how such purposes can be achieved. not extremely; this kind of sampling may also be
referred to as intensity sampling (Patton, 1990, pp.
182–183).
Researchers using scores on instruments as the
THE HOW OF COMBINATIONS criterion for purposeful sampling may wish to col-
lect more data (e.g., via interviews or observa-
Mixed-method studies entail concrete operations tions) from the chosen participants for the purpose
at the technique level of research by which “qual- of triangulation: that is, to discern whether a typi-
itative” and “quantitative” techniques are used to- cal, extreme, or intense case of something on a
gether and either remain distinct design compo- standardized test is also a typical, extreme, or in-
nents, or are explicitly integrated (Caracelli & tense case using other data collection techniques.
Greene, 1997). As shown in Figure 1, either qual- Or, researchers may use the criterion of scores
itative or quantitative approaches to sampling, (typical, extreme, or intense scores) for the pur-
data collection, and data analysis may prevail or pose of complementarity: that is, to find out more
have equal priority in a study, and both qualitative about what makes a case typical, extreme, or in-
1098240x, 2000, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1098-240X(200006)23:3<246::AID-NUR9>3.0.CO;2-H by Cochrane Philippines, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
COMBINING / SANDELOWSKI 249

FIGURE 1. Hybrid, combination, or mixed-method design templates.

tense. In the process of sampling for complemen- Random purposeful sampling. Another ex-
tarity, researchers will inevitably also obtain in- ample of the combined use of probability and pur-
formation on convergent validity and, thereby, poseful sampling is random purposeful sampling,
also achieve the purpose of triangulation. That is, which may also be used in design template 2 or 3.
in the process of obtaining fuller information on This sampling strategy is employed when there is
why persons scored as they did, they will also ob- a very large pool of potentially information-rich
tain information on whether persons look the same cases and no obvious reason to choose one case
on interview or observation as they did on the over another. For example, in a clinical trial of an
quantitative measure of a target phenomenon. Re- intervention to reduce pain with 500 people, 300
searchers will sample participants in scoring cate- of them scored as having less pain on a standard-
gories until the point of informational redundan- ized measure of pain, 150 scored as having as
cy; that is, until they have collected information much pain as they had before the treatment, and 50
from enough cases in each scoring category to al- scored as having more pain. These numbers are
low them to draw conclusions about the validity of too large for any purposeful sampling strategy ori-
the result (if they are seeking convergent validity), ented toward the intensive study of the particulars
or to elaborate on and clarify the result. Sampling of each case. Accordingly, cases can initially be
on the basis of scores is an especially useful strat- chosen from each of these three scoring groups
egy in clinical trials of interventions to validate or (the criterion for sampling here is again the scores)
clarify their effects on different participants. by assigning all the cases in each group a number
1098240x, 2000, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1098-240X(200006)23:3<246::AID-NUR9>3.0.CO;2-H by Cochrane Philippines, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
250 RESEARCH IN NURSING & HEALTH

from a random number table and then drawing illness; 1–2 cases of a husband caring for his wife,
them in turn. Each case drawn must meet the min- who suffers from a mortal and transmissible ill-
imum criterion in all purposeful sampling: name- ness; and, 1–2 cases of a husband caring for his
ly, that it is an information-rich case. And, only so wife, who suffers from a nonmortal and transmis-
many cases are drawn in each scoring group that sible illness. The researcher may also want to have
will permit researchers to make the kinds of infer- cases of wives caring for husbands, which show
ences they wish to make: for example, concerning the variations just itemized on illness mortality
convergent validity, or fuller description or expla- and transmissibility. In addition, the researcher
nation of cases. may want to include these variations for as many
Stratified purposeful sampling. Another kind of the other kinds of caregiving dyads previously
of combination of sampling techniques is stratified listed as s/he can. Or, the researcher may want to
purposeful sampling, where the researcher wants restrict sampling only to one kind of caregiving
to ensure that certain cases varying on preselected dyad (eg., parent–child caregivers) varying on the
parameters are included. This strategy can poten- parameters of member role and type of illness.
tially be used in any of the design templates shown In short, in stratified purposeful sampling, re-
in Figure 1. Although this kind of sampling is— searchers want to fill each sampling cell with 1-2
from a probability sampling standpoint—statisti- cases that exemplify the kinds and degree of vari-
cally nonrepresentative (Trost, 1986), it is, from a ation they surmise are relevant to understanding a
purposeful sampling standpoint, informationally target phenomenon, such as caregiving. This kind
representative. Each case represents a prespecified of sampling typically involves empirical, as op-
combination of variables, the distinctive conflu- posed to theoretical, cases that represent combina-
ence of which is the focus of study. The researcher tions of demographic (e.g., age, sex, income, and
wants to explain how these variables come to- education) and other low-inference variables. Low-
gether to make a case the case that it is. inference variables are variables concerning which
For example, a researcher studying caregiving there is likely to be consensus about what they are.
may wish to ensure that s/he has 1-2 cases that That is, there is likely to be no disagreement on
minimally to maximally vary on four parameters: whether a case represents a wife caring for her sick
type of caregiving dyad, role of each member of husband. In contrast, an empirical case of Mrs.
dyad, mortality of illness, and transmissibility of Jones caring for her husband, Mr. Jones, who is
illness. Caregiving dyads include parents caring dying from cancer, might also represent a theoret-
for their children, adult or minor children caring ical case of “reluctant caregiving,” as opposed to
for their parents, husbands caring for their wives, “engaged caregiving.” Although researchers can
wives caring for their husbands, heterosexual part- use a stratified purposeful sampling strategy to fill
ners caring for each other, male and female ho- sampling cells with such theoretical cases, they
mosexual partners caring for each other, brothers must first have made the case for such theoretical
caring for sisters, sisters caring for brothers, cases. Figure 2 illustrates a stratified purposeful
grandparents caring for grandchildren, grandchil- sampling plan restricted to one kind of caregiving
dren caring for grandparents, friends caring for dyad (parent–child), but varying on member role
friends, strangers caring for strangers, paid care- and type of illness. A total sample size of 16 would
givers caring for patients, and so on. In each of be required to fill each sampling cell with one
these dyads, one or the other member can be the case. Figure 2 also illustrates why certain sam-
caregiver or the cared-for; accordingly, wives care pling cells may be difficult or impossible to fill;
for or are cared for by husbands. Finally, illnesses few parent–child dyads will be comprised of a mi-
can be classified as mortal or nonmortal, and nor child caring for her/his own minor ill child.
transmissible or nontransmissible.
Accordingly, a researcher may want to have the Combining Data Collection Techniques
most variation s/he can achieve on each of these
four parameters, or limit variation on one or more Another set of concrete operations at the technique
of these parameters. In a maximally varied form of level of research entail the combined use of data
purposeful stratified sampling, the researcher might collection techniques that are commonly (but not
want to have 1–2 cases of every combination of necessarily) associated with either qualitative or
variations s/he can find. The researcher may want quantitative research, such as open-ended and un-
to have 1–2 cases of a husband caring for his wife, structured interviewing and structured question-
who suffers from a mortal and nontransmissible naires, respectively. (Some would argue that all
illness; 1–2 cases of a husband caring for his wife, data collection techniques in human subjects re-
who suffers from a nonmortal and nontransmissible search, including instruments, are “qualitative” in
1098240x, 2000, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1098-240X(200006)23:3<246::AID-NUR9>3.0.CO;2-H by Cochrane Philippines, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
COMBINING / SANDELOWSKI 251

FIGURE 2. Illustration of stratified purposeful sampling plan.

that they involve verbal data, which are only later tests were resolved by the analysis of ethnographic
transformed into numbers.) Researchers’ viewing interviews, which provided them the opportunity
positions will influence how they use these tech- to deepen their understanding of the use of anti-
niques. For example, for many researchers in a malarial chemoprophylaxis during pregnancy. This
positivist viewing position, data collection tech- study also shows the inherently multi-technique
niques vary in the degree to which they yield ob- nature of ethnographic studies.
jective data. Observations of behavior are gener-
ally thought to be more objective than self-reports The Varied Uses of Instruments
of behavior. When a target phenomenon can be ob-
served, observation is often the criterion measure Instruments can be used in combination studies to
against which self-report is judged. Accordingly, fulfill a variety of objectives. They can be used to
researchers in a positivist viewing position will of- provide fuller description of cases in areas sug-
ten seek “more objective” measures to evaluate gested by interview or observation data, as, for ex-
the validity of “more subjective” measures. More- ample, when interview data suggest that partici-
over, whenever there is a discrepancy between pants are depressed and the researcher decides to
what participants do and say they do, what obser- administer a depression inventory. Each partici-
vers see participants doing is generally considered pant’s score may be compared to the normative
a more accurate reflection of reality than self-report. score for that instrument or to the scores of other
The self-report is typically called into question. participants in the study. Instruments are used here
In contrast, from a constructivist viewing posi- to make case-bound or idiographic generalizations
tion, there is no hierarchy of data collection tech- —that is, generalizations about each participant in
niques whereby one technique is judged to yield the study—not to make nomothetic generaliza-
more objective (or more accurate or truer) data tions, or generalizations from the study sample to
than another. If the results from two data collec- populations. This use of instruments exemplifies
tion techniques do not converge, these results are design template 1 shown in Figure 1, where the de-
treated as interpretive opportunities: either to cision to use an instrument is made on analytic
show that no true discrepancy exists or to suggest grounds developed from data collected in a quali-
the phenomenon that accounts for the apparent tative study. This use also exemplifies the “develop-
discrepancy. For example, Silverman (1993) dem- ment” purpose for mixed-method research, where-
onstrated how a discrepancy between parents’ ac- by the results of using one kind of data collection
counts of their behavior and observations of their technique informs or guides the use of another
behavior in an examining room could be account- kind. Researchers may also use instruments for
ed for without resorting to judgments about these description (as opposed to statistical inference)
parents as unreliable informants. Indeed, he ad- concurrently with interviews or observations for
vised against the kind of “simple-minded ‘trian- the purpose of complementarity, as exemplified in
gulation’” whereby one kind of data is used sim- design template 1a, or as a development bridge be-
ply to corroborate or refute the results of another tween the qualitative exploration of a target phe-
without attention to the “embedded, situated na- nomenon and the further explanation of this phe-
ture of accounts” (p. 200). Helitzer-Allen and nomenon, as shown in design template 6. The
Kendall (1992) described how “discrepancies” be- development bridge here—the instrument—di-
tween data derived from surveys and laboratory rects researchers toward whom they will sample
1098240x, 2000, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1098-240X(200006)23:3<246::AID-NUR9>3.0.CO;2-H by Cochrane Philippines, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
252 RESEARCH IN NURSING & HEALTH

and what data they will collect in the second qual- the symptoms listed was “headache.” One waiting
itative portion of a study. father spontaneously mentioned to me that he had
Another use of instruments that fulfills the experienced a headache from hitting his head on a
purpose of development is to guide purposeful pipe doing a home repair, but, on the symptom in-
sampling. The results of instruments can direct re- ventory, did not mark headache as one of his
searchers more precisely to the kinds of partici- symptoms. As this man explained it, he surmised
pants they may wish to recruit and the nature of in- that we were not interested in symptoms that de-
formation they will want to obtain from them. As rived from home repairs or other events apparent-
described previously, instrument scores can be the ly unconnected to waiting for a child. What this
basis for sampling in a follow-up qualitative study. man clarified for me was all of the judgments par-
Results from a survey can direct researchers to- ticipants make when they encounter any item on
ward the most fruitful variables and associations an instrument of which researchers are complete-
to examine further both qualitatively, as shown in ly unaware. Accordingly, researchers may want to
design template 3 in Figure 1, or first qualitatively explicitly use instruments as data elicitation de-
and then again, quantitatively, as shown in design vices to determine exactly what participants see
template 7 in Figure 1. Ornstein and his colleagues and, therefore, respond to in each item. Because
(1993) initiated focus groups after learning from participants may also offer reasons for responding
the results of a telephone survey that there were as they did, in addition to explanations for how
three groups of “nonresponders” to a reminder they and persons like them are likely or ought to
letter for cholesterol screening. The focus groups respond, researchers may obtain information not
were organized to elicit more information concern- only on the content validity of an instrument, but
ing the lack of response. This information could also on its construct validity.
then serve as the basis for revising the reminder Instrument scores can also be used for qualita-
protocol and then qualitatively or quantitatively tive profiling. Qualitative profiling, which is dis-
evaluating its effectiveness in reducing the num- cussed further in the section on “qualitizing,” en-
bers of nonresponders. For example, the new re- tails theoretically grouping or typing participants
minder protocol could be tested in a clinical trial according to their scores on two or more instru-
and the scores on the instruments used to appraise ments. These typologies can be used to guide the-
the protocol could then be used as the basis for cri- oretical sampling in grounded theory studies de-
terion sampling to further explain those scores. signed to develop theory.
This combined example illustrates how a quan .
Qual study (survey>focus group) can lead to a Combining Data Analysis Techniques
Quan . qual (experiment . criterion sampling 1
interviews) study. Together, these studies can be Qualitative and quantitative data sets can be linked,
represented as quan.Qual. Quan.qual. preserving the numbers and words in each data set.
Instruments can also be used as elicitation de- Or, these data can be transformed to create one data
vices in interviews concerning both the target phe- set, with qualitative data converted into quantita-
nomenon and the instrument itself. For example, a tive data, or quantitative data converted into quali-
researcher may use participants’ responses on a tative data (Caracelli & Greene, 1993).
depression inventory to trigger more thoughts and Linking the results of qualitative and quantita-
feelings about depression in individual interview tive analysis techniques is accomplished by treat-
sessions. Participants often require some assis- ing each data set with the techniques usually used
tance to articulate inchoate thoughts or to speak with that data; that is, qualitative techniques are
the unspeakable. The use of an instrument as an used to analyze qualitative data and quantitative
elicitation device can serve this purpose, just as techniques are used to analyze quantitative data.
the use of projective techniques (e.g., Ornstein et For example, constant comparison, qualitative con-
al., 1993) can help participants language, and fo- tent, and narrative analysis techniques are used to
cus on, a target experience. analyze interview data, whereas one or more statis-
Researchers may also ask participants to com- tical techniques are used to analyze data from in-
ment specifically on their view of each item on an struments. The results of the qualitative analysis
instrument in order to appraise its content and con- of qualitative data and of the quantitative analysis
struct validity. In the infertility study mentioned of quantitative data are then combined at the inter-
previously (Sandelowski et al., 1992), we used a pretive level of research, but each data set remains
symptom inventory to appraise the type and in- analytically separate from the other.
tensity of symptoms women and men were expe- In contrast to the process of linking data are
riencing as they awaited birth or adoption. One of treatments of data that transform one kind of data
1098240x, 2000, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1098-240X(200006)23:3<246::AID-NUR9>3.0.CO;2-H by Cochrane Philippines, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
COMBINING / SANDELOWSKI 253

into another kind to create one data set. Tashakkori liability measures to validate them. In contrast, a
and Teddlie (1998, p. 126) referred to the conver- constructivist using narrative methodology would
sion of qualitative data into quantitative data as emphasize the meaningfulness of the data and
“quantitizing,” and to the conversion of quantita- would treat those data as inherently revisionist
tive data into qualitative data as “qualitizing.” and, thus not amenable to reliability testing.
Quantitizing. Quantitizing refers to a process Quantitative treatments of qualitative data can
by which qualitative data are treated with quantita- also be used to extract more information from
tive techniques to transform them into quantitative qualitative data, and to confirm researchers’ im-
data. The researcher must first reduce verbal or vi- pressions from these data. For example, from inter-
sual data (e.g., from interviews, observations, arti- view data with infertile couples, we suspected an
facts, or documents) into items, constructs, or vari- association between physicians’ encouragement
ables that are intended to mean only one thing and to have an amniocentesis performed and couples’
that can, therefore, be represented numerically. One decision to undergo the procedure (Sandelowski,
of the most commonly used examples of this pro- Harris, & Holditch-Davis, 1991). Accordingly, in
cess (and of design template 4 in Figure 1) is the cre- order to validate our impression, we created a data
ation of items for an instrument from interview display to discern the congruence between deci-
data. Although the researcher’s intention is to “pre- sion and advice by counting the numbers of cou-
serve qualitative meaning” in the development of ples who were: (a) encouraged to have the proce-
instruments (Fleury, 1993), any one item can only dure and had it (n 5 7); (b) encouraged to have the
have one meaning to function well psychometrical- procedure but did not have it (n 5 3); (c) not en-
ly. A valid item in an instrument cannot be mean- couraged or discouraged to have the procedure
ingful. That is, it cannot be full of meaning; it can- and did not have it (n 5 7); and, (d) not encour-
not mean different things at the same or different aged or discouraged to have the procedure but did
times to the same or different people. have it (n 5 3). The display visually confirmed our
Another example of a quantitative transforma- impression of an association between the decision
tion aimed at producing a one-meaning unit is the to have the procedure and the physician’s advice.
reduction of narrative data to a variable that can be We then analyzed these data using Fisher’s exact
correlated with other variables. In a study exem- probability test, which showed a nonsignificant
plifying mixed-method design template 4 in statistical relationship (N 5 20; p 5 .078), which
Figure 1, in which quantitative methods prevail might reach significance in another study with a
but follow and depend on the results of qualitative larger sample size.
methods, Borkan, Quirk, and Sullivan (1991) eli- Qualitizing. Qualitizing refers to a process
cited narrative data from elders concerning how by which quantitative data are transformed into
they viewed the hip fractures they had suffered. qualitative data. As in quantitizing, qualitizing can
Using a form of narrative analysis, they determined be used to extract more information from quanti-
that there were two major narrative emplotments tative data, or to confirm interpretations of it.
of hip fracture: the mechanical and the organic. An example of this process is the use of scores
They conducted a series of reliability tests to en- on instruments to profile participants—to create
sure that each narrative was consistently classified verbal portraits or typologies of them—around
as mechanical or organic. They then grouped el- target phenomena. Tashakkori and Teddlie (1998,
ders according to whether they had told a largely pp. 130–133) described five kinds of narrative or
mechanical or organic story about their hip frac- qualitative profiling: modal, average, comparative,
ture, and then conducted a correlation study to de- normative, and holistic. A modal profile is a verbal
termine whether and in what direction the emplot- description of a group of participants around the
ments-turned-into-variables—that is, mechanical most frequently occurring attributes. For example,
narrative and organic narrative—predicted func- if most of the participants in a group are in their
tional outcomes. They found that the organic sto- 50s, the group can be described as middle-aged. If
ry predicted poorer outcomes than the mechanical most of the participants in a group score in a cer-
story. tain range on a depression inventory, they can be
This study is an especially good illustration of described as mildly or severely depressed. Simi-
the axiom that methods are not uniformly linked larly, an average profile is a verbal description of
to paradigms. Borkan and his colleagues used a a group of participants around the mean of an at-
narrative methodology for positivist ends. They tribute. Although not usually identified in research
sought to create a predictor variable by reducing reports as qualitative profiling, this kind of quali-
the interview data they had obtained to units with tizing is commonly done to describe samples and
only one meaning and then appropriately used re- to interpret research results.
1098240x, 2000, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1098-240X(200006)23:3<246::AID-NUR9>3.0.CO;2-H by Cochrane Philippines, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
254 RESEARCH IN NURSING & HEALTH

A comparative profile is a verbal description been using combinations of qualitative and quan-
based on the comparison of participants to each titative techniques, but the qualitative techniques
other on one or more sets of scores. In contrast, a have not been featured, have been inappropriately
normative profile is a verbal description based on or inadequately used or explained, or have been
the comparison of participants’ scores to the nor- explicitly minimized. Sutton (1997) lauded the
mative scores for one or more instruments. Both “virtues of closet qualitative research” whereby
of these kinds of qualitative profiling depend on researchers “conceal or downplay” (p. 97) the use
the results of quantitative cluster analyses, which of qualitative techniques in order to have their re-
entail a set of statistical techniques to identify ho- search reports published!
mogeneous groups of subjects, or distinctive sub- The combined use of qualitative and quantita-
groups in which subjects may be placed because tive techniques will inevitably be informed by the
they are more similar to each other than to subjects researcher’s viewing position, which shapes what
in other subgroups. For example, Rothert and her techniques will be combined, and how and why
colleagues (1990) identified four groups of women they are combined. Accordingly, if researchers
on the basis of their responses to eight scenarios want to combine “stories and numbers. . . without
concerning hormone replacement therapy. That is, compromise” (Ford-Gilboe, Campbell, & Berman,
they created a typology of information use that 1995), use “numbers and words” in a “shamelessly
showed how differently people can respond to the eclectic” manner (Rossman & Wilson, 1994), and
same sets of information, and that these differences ease the “uneasy alliance” (Buchanan, 1992) be-
could be clustered in ways that distinguished tween qualitative and quantitative methods, they
women from each other. After they quantitatively must have a clear view of their viewing positions
identified these groups, they interviewed three and what dynamic mixes they suggest or permit.
women from each of these groups for further clar- Researchers must also resist the “‘mix and
ification and to confirm their typology. Their work match syndrome’” (Leininger, 1994, p. 103) that
is an example of design template 2 in Figure 1. can result in a “qualitative quagmire” (Barbour,
Qualitizing entails further studying these theo- 1998) and that mandates that all research be
retical groups to explicate the features of members mixed-method research. As Chen (1997) suggest-
in each group that make them like each other and of ed, mixed-method research is in danger of becom-
the groups themselves that make them different ing the new ideology. Mixed-method research
from each other, and to choose a name that will cap- should never be used because of the misguided as-
ture these features. This kind of investigation lends sumptions that more is better, that it is the fash-
itself especially well to subsequent grounded theory ionable thing to do, or, most importantly, that qual-
study, as the clusters can provide the basis for theo- itative research is incomplete without quantitative
retical sampling and for typology development. research (Morse, 1996). Indeed, qualitative tech-
Researchers can then validate these groupings with niques have been used to “salvage” quantitative
further qualitative study or quantitative study, as studies (Weinholtz, Kacer, & Rocklin, 1995). The
exemplified in design templates 5 or 7. “completeness” of any individual study, no matter
Finally, a holistic profile is a verbal description what kind it is, must be judged without resorting
based on impressions rather than specific attribut- to methodological fads or fetishes.
es or scores. Holistic profiles may also be com-
prised of various combinations of modal, mean,
comparative, and normative profiles. REFERENCES

Annells, M. (1996). Grounded theory method: Philo-


CONCLUSION sophical perspectives, paradigm of inquiry, and post-
modernism. Qualitative Health Research, 6, 379–393.
Mixed-method research is a dynamic option for Barbour, R.S. (1998). Mixing qualitative methods:
expanding the scope and improving the analytic Quality assurance or qualitative quagmire? Qualita-
power of studies. When done well, mixed-method tive Health Research, 8, 352–361.
studies dramatize the artfulness and versatility of Berman, H., Ford-Gilboe, M., & Campbell, J.C. (1998).
Combining stories and numbers: A methodologic ap-
research design. Mixed-method research opera- proach for a critical nursing science. ANS: Advances
tionally includes an almost limitless array of com- in Nursing Science, 21(1), 1–15.
binations of sampling, and data collection and Borkan, J.M., Quirk, M., & Sullivan, M. (1991). Find-
analysis techniques, only a few of which could be ing meaning after the fall: Injury narratives from el-
described here. Indeed, closer examination of re- derly hip fracture patients. Social Science & Medi-
search reports will show researchers have long cine, 33, 947–957.
1098240x, 2000, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1098-240X(200006)23:3<246::AID-NUR9>3.0.CO;2-H by Cochrane Philippines, Wiley Online Library on [25/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
COMBINING / SANDELOWSKI 255

Buchanan, D.R. (1992). An uneasy alliance: Combining Morse, J.M. (1991). Approaches to qualitative-quanti-
qualitative and quantitative research methods. Health tative methodological triangulation. Nursing Re-
Education Quarterly, 19, 117–135. search, 40, 120–123.
Caracelli, V.J., & Greene, J.C. (1993). Data analysis Morse, J.M. (1996). Is qualitative research complete?
strategies for mixed-method evaluation designs. Edu- Qualitative Health Research, 6, 3–5.
cational Evaluation and Policy Analysis, 15, 195–207. Ornstein, S.M., Musham, C., Reid, A., Jenkins, R.G.,
Caracelli, V.J., & Greene, J.C. (1997). Crafting mixed- Zemp, L.D., & Garr, D.R. (1993). Barriers to adher-
method evaluation designs. In J.C. Greene & V.J. Cara- ence to preventive services reminder letters: The pa-
celli (Eds.), Advances in mixed-method evaluation: tient’s perspective. Journal of Family Practice, 36,
The challenges and benefits of integrating diverse par- 195–200.
adigms (pp. 19–32). San Francisco: Jossey-Bass. Patton, M.Q. (1990). Qualitative evaluation and re-
Chen, H. (1997). Applying mixed methods under the search methods (2nd ed.). Newbury Park, CA: Sage.
framework of theory-driven evaluations. In J.C. Pearce, J.C. (1971). The crack in the cosmic egg: Chal-
Greene & V.J. Caracelli (Eds.), Advances in mixed- lenging constructs of mind and reality. New York:
method evaluation: The challenges and benefits of in- Washington Square Press.
tegrating diverse paradigms (pp. 61–72). San Fran- Rossman, G.B., & Wilson, B.L. (1994). Numbers and
cisco: Jossey-Bass. words revisited: Being “shamelessly eclectic.” Qual-
Fleury, J. (1993). Preserving qualitative meaning in in- ity & Quantity, 28, 315–327.
strument development. Journal of Nursing Measure- Rothert, M., Rovner, D., Holmes, M., Schmitt, N., Ta-
ment, 1, 135 –144. larczyk, G., Kroll, J., & Gogate, J. (1990). Women’s
Ford-Gilboe, M., Campbell, J., & Berman, H. (1995). use of information regarding hormone replacement
Stories and numbers: Coexistence without compro- therapy. Research in Nursing & Health, 13, 355–366.
mise. ANS: Advances in Nursing Science, 18, 14–26. Sandelowski, M. (1995). Triangles and crystals: On the
Greene, J.C., & Caracelli, V.J., (Eds.). (1997a). Ad- geometry of qualitative research. Research in Nurs-
vances in mixed-method evaluation: The challenges ing & Health, 18, 569–574.
and benefits of integrating diverse paradigms. San Sandelowski, M., Harris, B.G., & Holditch-Davis, D.
Francisco: Jossey-Bass. (1991). Amniocentesis in the context of infertility.
Greene, J.C., & Caracelli, V.J. (1997b). Defining and Health Care for Women International, 12, 167–
describing the paradigm issue in mixed-method eval- 178.
uation. In J.C. Greene & V.J. Caracelli (Eds.), Ad- Sandelowski, M., Holditch-Davis, D., & Harris, B.G.
vances in mixed-method evaluation: The challenges (1992). Using qualitative and quantitative methods:
and benefits of integrating diverse paradigms (pp. 5 – The transition to parenthood of infertile couples. In
17). San Francisco: Jossey-Bass. J.F. Gilgun, K. Daly, & G. Handel (Eds.), Qualitative
Greene, J.C., Caracelli, V.J., & Graham, W.F. (1989). methods in family research (pp. 301–322). Newbury
Toward a conceptual framework for mixed-method Park, CA: Sage.
evaluation designs. Educational Evaluation and Pol- Silverman, D. (1993). Interpreting qualitative data:
icy Analysis, 11, 255 –274. Methods for analyzing talk, text and interaction. Lon-
Guba, E.G., & Lincoln, Y.S. (1994). Competing para- don: Sage.
digms in qualitative research. In N.K. Denzin & Y.S. Sutton, R.I. (1997). The virtues of closet qualitative re-
Lincoln (Eds.), Handbook of qualitative research search. Organization Science, 8, 97–106.
(pp. 105 –117). Thousand Oaks, CA: Sage. Swanson, S.C. (1992). A cross-disciplinary application
Helitzer-Allen, D.L., & Kendall, C. (1992). Explaining of Greene, Caracelli and Grahams’s conceptual
differences between qualitative and quantitative data: framework for mixed method evaluation. Unpub-
A study of chemoprophylaxis during pregnancy. lished doctoral dissertation, Georgia State Universi-
Health Education Quarterly 19, 41– 54. ty, Atlanta, GA.
Heron, J., & Reason, P. (1997). A participatory inquiry Tashakkori, A., & Teddlie, C. (1998). Mixed methodol-
paradigm. Qualitative Inquiry, 3, 274 –294. ogy: Combining qualitative and quantitative ap-
Leininger, M. (1994). Evaluation criteria and critique in proaches. Thousand Oaks, CA: Sage.
qualitative research studies. In J.M. Morse (Ed.), Trost, J.E. (1986). Statistically non-representative strat-
Critical issues in qualitative research methods ified sampling: A sampling technique for qualitative
(pp. 95 –115). Thousand Oaks, CA: Sage. studies. Qualitative Sociology, 9, 54–57.
Miles, M.B., & Huberman, A.M. (1994). Qualitative Weinholtz, D., Kacer, B., & Rocklin, T. (1995). Sal-
data analysis: An expanded sourcebook (2nd ed.). vaging quantitative research with qualitative data.
Thousand Oaks, CA: Sage. Qualitative Health Research, 5, 388–397.
Morgan, D.L. (1998). Practical strategies for combining Wolfer, J. (1993). Aspects of “reality” and ways of
qualitative and quantitative methods: Applications to knowing in nursing: In search of an integrating para-
health research. Qualitative Health Research, 8, 362– digm. Image: Journal of Nursing Scholarship, 25,
376. 141–146.

You might also like