sign in sign up
<<
Home , Postpositivism, Qualitative research

View metadata, and similar papers at core.ac.uk brought to you by CORE

provided by AUT Scholarly Commons

Copyright © eContent Management Pty Ltd. Contemporary Nurse (2006) 23: 3–11.

Mixed methods for the novice researcher

ABSTRACT Mixed methods research is becoming increasingly popular in the health and disciplines.The aim of this article is to give an overview of the Key Words varieties of mixed methods designs.We begin by situating mixed methods research in the context of a paradigmatic framework which assists a researcher mixed methods; qualitative in making decisions concerning the design of their study.Although the most research; commonly used mixed methods designs are underpinned by positivist/ quantitative postpositivist assumptions, the combination of qualitative and ; methods can be used within any research paradigm. paradigms; Received 24 January 2006 Accepted 27 July 2006 CN

BARBARA M GRANT Senior Lecturer LYNNE S GIDDINGS Centre for Associate Professor Professional School of Nursing Development Auckland University The University of of Auckland Auckland, Aotearoa Auckland, Aotearoa New Zealand New Zealand

INTRODUCTION: either unable to answer the ANYTHING GOES? or trying to push contradictory data into a t first glance, mixed methods appears compromising fit. In our experience, health and A to offer an ‘anything goes’ approach to social science students undertaking research for research. Not so. There is always a taken-for- the first time often struggle with the inevitable granted, and usually unacknowledged, world- complexities and messiness of the research view with underlying assumptions that guides design process.Yet the choices we make about the choice of methods to be mixed and how the methods are important because particular data will be used. If this or paradigm methods will close down or open up research is not made explicit during the research possibilities in quite distinctive ways; they will process, a researcher may find themselves lost, allow some questions to be explored but not

Volume 23, Issue 1, October 2006 CN 3 CN Lynne S Giddings and Barbara M Grant

others. When undertaking a mixed methods combining from different para- study careful consideration needs to be given to digms) without getting into contradiction. the assumptions underpinning the research We argue that postpositivism is a shift from approach so there is congruence between the within the positivist paradigm. It developed dur- chosen methods and the research question. A ing the 1960s, largely from the increasing fundamental assumption underpinning this arti- recognition without and within science of the ide- cle is that mixed methods is a research tool ological and practical limitations of certain rather than a in its own right. designs and strategies.The prefix ‘post’, when attached to words like , , A researcher’s worldview and colonialism for example, indicates a further or paradigm development of the original concept, but one In this paper we focus on the most commonly that is fundamentally critical of it. So postposi- used approach to mixed methods research, a tivism continues most of the key philosophical combination of qualitative and quantitative assumptions of positivism but in a changed or methods for collecting and analysing data under- more moderate form. For example, a core pinned by postpositivist assumptions.We view fundamental positivist assumption is that of postpositivism as an extension of the traditional determinism, the belief that effects have a scientific worldview or paradigm known as posi- determinable cause and actions have predictable tivism (as we will shortly explain). In our view outcomes. Postpositivists maintain this assump- (described in detail in Grant & Giddings, 2002), tion in a modified form: rather than assuming a the main health and social science research linear process of cause and effect, they perceive paradigms are four: the positivist/scientific, outcomes as the result of a complex array of interpretivist/constructivist, radical/critical and causative factors that interact with each other. poststructural. A researcher’s paradigm reflects Mixed methods researchers are not always their beliefs about what reality is (ontology), aware of the postpositivist underpinning of their what counts as (), how studies. By omission, their work may reflect an one gains knowledge (methodology), and the assumption of being paradigm free or they may values one holds (). The first three make unsupported claims of creating the ‘best terms may be scarily familiar to you but the last of both worlds’ by incorporating other para- is likely to be new.Yet the axiological position- digms such as interpretive or radical/critical. ing of a researcher is often the determining fac- One of us has described such mixed methods tor in the research decision-making process. For studies as ‘positivism dressed in drag’ (Giddings instance, a person who has strong values on 2005: 195). issues of social justice and equity is likely to be In what follows we begin by clarifying the drawn to the radical/critical paradigm because terms methodology/methods and qualitative/ it focuses on social action and .The quantitative.We then offer some guidelines on emergence of Kaupapa Maori research (indige- how to decide when to use a mixed methods nous within Aotearoa New Zealand) may be design, followed by a description of the various understood in part as driven by profound designs we consider most useful in health axiological differences embedded in different research. Illustrative examples from nursing cultures.Taken together, paradigm assumptions research studies and references for further read- and beliefs indicate the proper kind of ing are provided along the way.Although mixed researcher–researched relationship. Important- methods research is largely located in post- ly, paradigms are incommensurate, that is you positivism, we suggest that it can be used by cannot easily work across them (for instance by researchers situated within other paradigms.

4 CN Volume 23, Issue 1, October 2006 Mixed methods research for the novice researcher CN

WHAT IS ‘MIXED’ IN MIXED In this sense the methods are ‘a-theoretical and METHODS RESEARCH? a-methodological’ (Sarantakos 1998: 34) and There is some debate over whether or not therefore can be mixed.Taking the blood testing mixed methods research is itself a research par- example above, the researcher could use the adigm or methodology. In our view, the most methods of interviewing the par- commonly used mixed methods designs are ticipants, measuring blood test results, and simply an approach to research that mixes quali- observing procedure, depending on their question. tative and quantitative methods.To understand To decide sensibly, the researcher would have to this, it is helpful to be clear about the difference identify their research question first. between methodology and methods, and the terms qualitative and quantitative. DECIDING TO USE MIXED Methodology is a more abstract term than METHODS RESEARCH methods and refers to the theoretical assump- Mixed methods research has a range of tions and principles that underpin a particular strengths. It is particularly useful in , eval- research approach. It guides how a researcher uation, and (Patton 2002) because frames the research question and decides on it has a broader focus than single method design what process and methods to use. Methods, in and gathers more information in different contrast, are much more concrete and practical modes about a phenomenon. It can also give – they are the tools for collecting and analysing insight into complex social phenomena such as data.The methods a researcher chooses need to family violence or anorexia nervosa by produc- fit with the research question. For example, if ing findings that illuminate that complexity.Yet your research question was ‘what is it like for another strength of mixed methods design is someone to have a blood test?’ (their experi- that the breadth of findings can bring value to ence), but then you focused your study on the the research process itself by highlighting the results of the blood tests (measurements) or particular shortcomings in each of the methods how the blood specimens were taken (observa- used and compensating for them. When the tions of technique), there would be a lack of fit. findings are contradictory, they can reveal The data you gathered would not allow you to researcher assumptions that would not other- address the research question. wise have been known or the constraints and The terms quantitative and qualitative are biases of ways of measuring or interpreting commonly used to describe both the methods something (refer to Exemplar 1 below for an and the methodologies used in health research. example of this). However, what makes this Confusing? Yes! Historically, quantitative most attractive to health practitioners is has been viewed as synonymous with positivism its , that is, its usefulness in the clin- and qualitative with interpretivism – hence the ical setting to collect comprehensive informa- association with methodology. Some writers tion about a phenomenon that can then guide consider the terms to refer to two research par- decisions about practice. Examples of such phe- adigms in and of themselves (Blaxter et al. nomena include: hospital admission and dis- 2001) or, at the other extreme, as terms to charge procedures; introduction of new merely describe forms of data: quantitative data processes, procedures or techniques; and the being numbers and statistics, and qualitative responses of clients to being asked about vio- being words and .We argue that the lence in their lives when attending an accident two terms most usefully describe different and emergency or health clinic. ‘types of methods’ (Guba & Lincoln 1994: 105) All research approaches have their limitations that may be used for data collection and . – so too mixed methods. Generally this design

Volume 23, Issue 1, October 2006 CN 5 CN Lynne S Giddings and Barbara M Grant

takes more time, both at the beginning for pre- qualitative method collect the most perti- planning and negotiation (because of the mix of nent data? This decision and the way you decide researcher skills needed) and at the end for to analyse your data are most likely to reveal the coming to agreement as to how the findings fit underpinning research paradigm that will guide together (or not) and what they ultimately the research process. mean. For these reasons, we suggest graduate 6. How does the secondary data collection students – especially in masters by research pro- method complement the primary? What grammes – approach the use of mixed methods does it contribute to the purpose of your with caution. However, its usefulness is such study? that it may be the best design to answer your 7. How will you sequence the methods? What question.We suggest if you decide to carry out needs to be known first? a mixed methods study that you find a sponsor 8. Do you have the skills to collect and analyse or a community of practice with others under- both data sets? Do you need to consult a taking the same research approach. Make sure methodological expert or include a co- as well that your supervisor is supportive of this researcher to complement your skills? approach and, even better, experienced with 9. How much time do you have to carry using it. out this study? (This is very relevant for As noted earlier, most mixed methods research students who often have a very research reflects postpositivist assumptions. limited time-frame.) Mixed methods research, however, is certainly not confined to this paradigm. It can be located Once you have answered these questions, you within others. For example, feminist researchers will have thought through most of the issues have long shown creative flexibility in their that bear on how you will carry out your mixed approaches to collecting and analysing data, able methods study.You will have begun to realise to utilise quantitative methods in the service of that you may not have a free choice in this radical/critical or poststructuralist paradigms. matter – that strong personal preferences Deciding where to locate your mixed meth- (yours but also your supervisor’s), as well as ods research is a process requiring some thought perhaps even pressures from your clinical – and probably discussion with others. Here are practice area, all have to be dealt with. some questions that can guide your thinking, writing, reading and talking at this time: MAIN DESIGNS OF MIXED 1. What is your issue/problem and what do METHODS RESEARCH you want to know about it? (This is your Various typologies have been developed to help research question and you may refine it over understand the many design possibilities in time as you collect and analyse data.) mixed methods research (Creswell 2003; Morse 2. What is your personal research orientation: 1991; Tashakkori & Teddlie 1998). The one what values, attributes, abilities, and research offered here in Table 1 is based on those but skills do you bring to the study? departs from them too. In our typology (as in 3. What are the dominant research traditions Creswell 2003) there are two main designs: in your /s? sequential (➔) and concurrent (+). We have 4. What methods best suit your research pur- abbreviated quantitative to QUANT or quant pose (aim or objective) and context (institu- and qualitative to QUAL or qual. tional, social, cultural, political)? 5. Which method is going to be the primary Sequential mixed methods designs data gatherer? Would a quantitative or a In sequential designs, one method is used first,

6 CN Volume 23, Issue 1, October 2006 Mixed methods research for the novice researcher CN

TABLE 1: TYPOLOGY OF MIXED METHODS RESEARCH DESIGNS (DERIVED FROM CRESWELL 2003) Sequential Concurrent

Equal QUAL ➔ QUANT QUANT + QUAL QUANT ➔ QUAL QUAL + QUANT

Unequal QUAL ➔ quant QUAL + quant qual ➔ QUANT qual + QUANT QUANT ➔ qual QUANT + qual quant ➔ QUAL Quant + QUAL followed then by the other: QUANT ➔ QUAL to gather data from her students. It is or QUAL ➔ QUANT. In the analysis and inter- not until the presentation of the findings that it pretation phase, the data may be treated equally is clear that a sequential mixed methods design (as suggested by the notation above) or one set was used, with the data from both methods of data may be seen as secondary as suggested by treated equally (QUANT ➔ QUAL): Ironside use of lower-case letters in these notations: first sets out the student responses to various QUANT ➔ qual, quant ➔ QUAL, QUAL items on the questionnaire giving some of the ➔ quant, or qual ➔ QUANT.We have cho- statistical results in support of conclusions sen as our exemplars two QUANT ➔ QUAL made. The qualitative data are then presented studies to show the variety of ways such studies thematically. In her analysis she attempts can be designed, and one QUANT ➔ qual to integrate the findings from the two data study.The last combination is one of the more sources. The quantitative findings were in a commonly used sequential designs in health and number of instances incongruent and inconsis- social science research. articles tent with the qualitative findings. Most of the are used for the exemplars. discussion in the article arises from the mis- match between these findings.This outcome of a Sequential exemplar 1 mixed methods study is one of its strengths. QUANT ➔ QUAL (integrated), New questions are posed and new ways to summarised in Table 2 explore them become possible. Pamela Ironside’s (2003) title Trying something Sequential exemplar 2 new: Implementing and evaluating pedagogy QUANT ➔ QUAL (separated), using a multimethod approach, clearly flags summarised in Table 3 that more than one method was used in her evaluative study. Ironside used a pretest–posttest In contrast to Ironside’s work, Jane Koziol- questionnaire followed by semi-structured McLain and colleagues (2004) used a sequential

TABLE 2: SEQUENTIAL EXEMPLAR 1: QUANT ➔ QUAL (INTEGRATED) Trying something new: Implementing and evaluating narrative pedagogy Ironside (2003) using a multimethod approach

Design • Pretest – posttest questionnaire followed by semi-structured interviews were used to gather data from her students.

Comment • Attempts to integrate the findings from both data sources. • Mismatch between findings becomes basis for new thinking.

Volume 23, Issue 1, October 2006 CN 7 CN Lynne S Giddings and Barbara M Grant

TABLE 3: SEQUENTIAL EXEMPLAR 2: QUANT ➔ QUAL (SEPARATED) Koziol-McLain Prevalence of intimate partner violence among women presenting et al. (2004) to an urban adult and paediatric emergency care department

Design • First a questionnaire, followed by a semi-structured telephone with a sub-sample.

Comment • Shared primary objective but different research questions and different primary investigators. • Both parts of study treated equally but separately with the findings presented as stand-alone studies and published in different journals.

design but separated the quantitative and quali- people with visible disfigurement, does not indicate tative parts of their study.They researched the that it is a mixed methods study. Neither do the topic of addressing family violence in the health- key words. From the title, the reader may won- care setting and described their design as mixed der if it is an interpretive qualitative study. methods. There was one research team, one Although not directly stated, it is the abstract ethics proposal, and a primary objective – to that situates the study firmly in the postposi- create change. They carried out their study tivist paradigm with the qualitative component sequentially. First they studied the prevalence of clearly secondary.The words that clue you in violence among women seeking health care by are ‘establish extent and type of psychosocial administering a questionnaire. They then con- needs’ and ‘little is known about levels of dis- ducted semi-structured telephone interviews tress’ (2004: 443); these are phenomena that with a sub-sample of those women, inquiring can be measured and described with quantita- about what the experience of being asked ques- tive methods. The study used cross-sectional tions about family violence in the healthcare set- survey design followed by semi-structured ting was like and if it made a difference for them interviews to ‘generate further quantitative and and/or their children. Both parts of the study qualitative data about individual concerns, and were treated equally but separately.They had a satisfaction with the provision of care’ (2004: shared primary objective but different research 443).This is a sequential mixed methods design questions, different members of their team were in which the qualitative component comple- primary investigators, the findings were pre- ments the quantitative (QUANT ➔ qual): sented as stand-alone studies, and submitted to statistical results are reported in text and in different journals for publication. tables with various findings being echoed by the interview data (2004: 450). Sequential exemplar 3 Concurrent mixed methods designs QUANT ➔ qual (integrated), summarised in Table 4 In the concurrent design, both methods are used at the same time. In some studies, both are The title of Rumsey et al.’s (2004) article, of equal importance to answering the research Altered body image: Appearance-related concerns of question but, like sequential designs, most often

TABLE 4: SEQUENTIAL EXEMPLAR 3: QUANT ➔ QUAL (INTEGRATED) Rumsey et al. (2004) Altered body image: Appearance-related concerns of people with visible disfigurement

Design • Cross-sectional survey followed by semi-structured interviews. Comment • Statistical results are reported in text/tables with findings being echoed by the interview data.

8 CN Volume 23, Issue 1, October 2006 Mixed methods research for the novice researcher CN one set of findings is used to complement the and quantitative methods.Their aim was two- other, e.g. QUAL + QUANT or QUANT + fold: to understand the needs of families who QUAL. Sometimes they are used in separate were attending court for drug dependency rea- sub-studies, and the data is analysed separately sons, and to ‘determine mothers’ perception of and then compared together.This is the widely PHN [Public Health Nurse] strategies that could advocated triangulation method of data collec- help with the difficult task of reunifying with tion where one set of data is used to corrobo- their children’ (p. 61). Quantitative methods rate the findings from another. In other studies, such as reviewing court files complemented the one method is ‘nested’ (Creswell, 2003: 218) data gained through the qualitative methods of inside the other and is usually subordinate: interviewing a group of mothers (n = 4) and QUANT + qual or QUAL + quant. The observing mothers during court proceedings. value of the secondary data set to the research is Not too far into reading the paper, however, one that it offers a broader perspective than the pri- becomes aware of paradigmatic dissonance – mary data could offer alone. However, difficul- the language simultaneously reflects positivist ties arise when attempting to integrate the and interpretive positionings and there are evi- findings: the data sets are often not compatible dent contradictions.Which paradigm underpins for analysis and result ‘in unequal the study? The paper’s title, Public health nurse within a study’ (Creswell 2003: 219). Com- interventions for women in a dependency drug court, monly, the inequality produces a bias in favour gives the first clue that the study’s research par- of the quantitative data because of its concrete adigm is positivist/postpositivist. Its focus on attributes – as ‘hard’ data, it is able to be repre- interventions signals an assumption of cause and sented in numbers and statistics.We have cho- effect (determinism).This paradigmatic conclu- sen as our exemplars a QUAL + QUANT and sion is supported on reading further: a QUANT + qual study as they raise interest- 1. A is given followed by a ing design issues that are important to consider description of a conceptual framework when planning a concurrent mixed methods which positions the theoretical notions study.The QUANT + qual combination is the explored in the study (theory verification). most commonly used mixed methods design in 2. No reference is made to methodology nursing research, the field from which we have (postpositivist positioning is not made drawn the exemplars. explicit). 3. Throughout the paper women who were Concurrent exemplar 1 interviewed (n = 4) were referred to both QUAL + QUANT,summarised in Table 5 as ‘subjects’ and ‘participants’ (lack of clarity as to researcher–researched relation- Somervell, Saylor and Mao’s (2005) study illus- ship). trates the complementary nature of qualitative 4. The women ‘were asked the same questions

TABLE 5: CONCURRENT EXEMPLAR 1: QUAL + QUANT Somervell et al. (2005) Public health nurse interventions for women in a dependency drug court

Design • (reviewing files), interviewing and observation.

Comment • Mix of methods enables a rich description of research topic. • Dissonance between interpretivist and positivist positioning gives rise to unacknowledged contradictions.

Volume 23, Issue 1, October 2006 CN 9 CN Lynne S Giddings and Barbara M Grant

during the interview process’ (p. 64) – (an 1950s the design consist of a survey question- attempt to control for bias and variability). naire using measurable items to produce quanti- 5. Although in the methods section reference tative (hard) data with either some open-ended was made to purposive sampling, in the questions or a space for comments to enable the conclusion the authors of the study noted collection of qualitative (soft) data. Borjesson et that a limitation was that the women ‘were al. (2004) questionnaire consisted of 37 items not randomly selected’ (p. 64) (would have (demographic details, Likert scales and multiple enabled a more representative sample to choice questions) with space underneath each ensure generalizability of the findings). for respondents to write comments. An open- ended question asking for their views on soci- The study by Somervell et al. (2005) highlights ety’s contribution to supporting mothers in some of the strengths of combining qualitative their parenting role was included at the end. and quantitative methods (a broad focus and a The qualitative data was clearly being used to variety of data collection approaches enables the support the quantitative findings.This nestled gathering of rich descriptions of a phenome- positioning of the qualitative component of the non), but also some of the ‘messiness’ that can study is evident too in the way the findings are occur when the paradigmatic positioning is not presented; the quantitative results are given first acknowledged or clearly understood.The use of with explanatory, descriptive, or supportive qualitative methods with a dusting of interpre- quotes from the respondents’ written com- tive (qualitative) concepts in a research report ments. Unlike the Somervell et al. (2005) study do not make a positivist/scientific study inclu- there was paradigmatic congruence. Although sive of the interpretive paradigm (the ‘best of the methodology was not specifically named as both worlds’); rather the qualitative methods descriptive mixed methods, its descriptive are used in the service of the . nature was consistent within and between each As demonstrated in Somervell et al. (2005) section as was the use of postpositivist language. research report, if the paradigmatic assumptions are not made explicit, the ensuring analysis may Reflections on mixed methods contain contradictory statements that challenge typologies the theoretical and methodological rigour of the Our typology shares many similarities with research. Creswell’s (2003). The major difference bet- ween our view and his arises around his Concurrent exemplar 2 categories of sequential and concurrent trans- QUANT + qual (nestled),summarised in Table 6 formative mixed methods designs. The key feature of transformative designs according to The QUANT + qual nestled design used by Creswell is that they are underpinned by ‘a the- Borjesson, Paperin and Lindell (2004) in their oretical perspective’ (p. 216) that guides the study of maternal support during the first year research and is more important than the choice of infancy, is the classical mixed methods of methods. In our view all research is theoreti- design. Used by health professionals since the cally guided but, as we have said, it is common

TABLE 6: CONCURRENT EXEMPLAR 2: QUANT + QUAL (NESTLED) Borjesson et al. (2004) Maternal support during the first year of infancy

Design • Survey questionnaire that includes quantitative and qualitative items. Comment • Qualitative data clearly used to support quantitative.

10 CN Volume 23, Issue 1, October 2006 Mixed methods research for the novice researcher CN that positivists (including postpositivists) do not Positivism dressed in drag? Journal of Research in acknowledge this. For example, the positivist Nursing 11(3): 195–203. claim that there is an objective reality to be dis- Grant BM and Giddings LS (2002) Making sense covered is a theoretical claim about the nature of of methodologies:A paradigm framework for human perception and knowing. Creswell also the novice researcher, Contemporary Nurse 13(1): takes the position that is 10–28. ‘one distinct methodology’ (p. 217); in contrast Guba EG and Lincoln YS (1994) Competing we understand there to be many distinctive paradigms in qualitative research, in: Denzin NK and Lincoln YS (Eds), Handbook of Qualita- methodologies that are qualitative and that these tive Research, pp. 105–117,Thousand Oaks, CA, cross diverse paradigms. In our view, the idea of Sage. ‘transformative’ research points towards the Ironside P (2003) Trying something new: Imple- radical/critical paradigm that is centrally con- menting and evaluating narrative pedagogy cerned with producing research that will lead to using a multimethod approach, Nursing social change (Grant & Giddings 2002).Where Education Perspectives 24(30): 122–128. we do agree with Creswell is in his Koziol-McLain J, Gardiner J, Batty P,Rameka M, that mixed methods research can be used not Fyfe E and Giddings LS (2004) Prevalence of only for descriptive/explanatory research but intimate partner violence among women also for transformative/critical work. presenting to an urban adult and paediatric emergency care department, New Zealand CONCLUSION Medical Journal 117(1206): U1174. Miles MB and Huberman AM (1994) Qualitative The increasing complexity of social and health :An Expanded Sourcebook,Thousand care issues demands creative ways of investigat- Oaks CA, Sage. ing and finding solutions to myriad problems. In Patton MQ (2002) Qualitative Research and Evalua- its various forms, mixed methods research is tion Methods, 3rd edn,Thousand Oaks, CA, Sage. now accepted as valuable ‘real world research’ Robson C (2002) Real World Research:A Resource for (Robson 2002) because it offers a versatility of Social Scientists and Practitioner–Researchers, 2nd approach. Moreover, it well suits the practical edn, Oxford, Blackwell Publishers. focus of researching nursing practice and client- Rumsey N, Clarke A,White P,Wyn-Williams M centred care. Although mixed methods research and Garlick W(2004) Altered body image: has been captured by postpositivism, it can be Appearance-related concerns of people with an effective approach for researchers from all visible disfigurement, Journal of Advanced paradigms. Nursing 48(5): 443–453. Sarantakos S (1998) , 2nd edn, Basingstoke, Macmillan. References Somervell AM, Saylor C and Mao C (2005) Public Blaxter L, Hughes C and Tight M (2001) How to health nurse interventions for women in a research, 2nd edn, Philadelphia PA, Open dependency drug court, Public Health Nursing University Press. 22(1): 59–64. Borjesson B, Paperin C and Lindell M (2004) Tashakkori A and Teddlie C (Eds) (2003) Handbook Maternal support during the first year of in- of Mixed Methods in Social & Behavioral Research, fancy, Journal of Advanced Nursing 45(6): Thousand Oaks CA, Sage. 588–594. Wuest J, Merritt-Gray M and Ford-Gilboe M Creswell JW (2003) Research Design:Qualitative, (2004) Regenerating family: Strengthening the Quantitative,and Mixed Methods Approaches (2nd emotional health of mothers and children in edn),Thousand Oaks CA, Sage. the context of intimate partner violence, Giddings LS (2005) Mixed methods research: Advances in Nursing Science 27(4): 257–274.

Volume 23, Issue 1, October 2006 CN 11