Conducting Mixed Analyses: A general typology

Anthony J Onwuegbuzie
Department of Educational Leadership and Counseling, Sam Houston State University, Huntsville TX, United States of America

John R Slate
Department of Educational Leadership and Counseling, Sam Houston State University, Huntsville TX, United States of America

Nancy L Leech
School of Education and Human Development, University of Colorado; Health Sciences Center, Denver CO, United States of America

Kathleen MT Collins
College of Education and Health Professions, University of Arkansas at Fayetteville AR, USA

PP: 4

Abstract

In this article, we provide a typology of mixed analysis techniques, namely the Mixed Analysis Matrix, that helps researchers select a data analysis technique given the number of (a) data types collected (ie quantitative or qualitative; or quantitative and qualitative) and (b) analysis types used (ie quantitative or qualitative; or quantitative and qualitative)—yielding a 2x2 representation involving four cells that each contain specific analytical techniques, with two of these cells containing a total of 15 mixed analysis techniques.

Furthermore, we describe the fundamental principle of mixed analysis, describe the steps in a mixed analysis, and delineate the rationale and purpose for conducting mixed analyses. For each technique, readers are directed to published studies that serve as illustrative examples. Outlining the mixed-analysis techniques available for researchers hopefully will increase awareness of the number of choices for analyzing data from mixed studies.

Keywords

mixed research, mixed analysis, data analysis typology, mixed analysis matrix, parallel mixed analysis, concurrent mixed analysis, sequential mixed analysis

Article Text

The collection of quantitative and qualitative data within the same study — currently most commonly called mixed methods research (Tashakkori & Teddlie, 2003a) and more recently as mixed research (Johnson & Onwuegbuzie, 2004) - has taken place for many decades, with the first article in which the phrase mixed methods was used appearing in 1972 (Parkhurst, Lovell, Sprafka & Hodgins, 1972). Professionals in the world of program evaluation are well aware of the important role that mixed research has long played in evaluation contexts (eg National Science Foundation, 2002). Formal techniques, however, for analyzing both quantitative and qualitative data within the same framework - a process known as mixed data analysis or, more simply, as mixed analysis - recently have emerged (eg Bazeley, 2006). With the exception of a handful of published works (eg Bazeley, 2003; Caracelli & Greene, 1993; Chi, 1997; Greene, Caracelli & Graham, 1989; Li, Marquart & Zercher, 2000; Onwuegbuzie & Leech, 2004; 2006; Onwuegbuzie & Teddlie, 2003; Sandelowski, 2000; 2001), over the last decade or more, relatively little specific information is provided as to how to analyze conduct mixed analyses.

Even in the various recent special issues on mixed research (eg Gorard & Smith, 2006; Onwuegbuzie & Leech, 2006), no article was devoted to mixed analyses. As such, in this article, the authors’ goal is to provide a typology of mixed-analysis techniques. We believe that such a typology is needed because it:

1. Assists in providing the mixed research field with a flexible organizational data-analysis structure
2. Helps to provide more credibility to the field of education in general and the social and behavioral sciences in particular by providing examples of mixed analyses that are markedly different than statistical analyses alone or qualitative analyses alone
3. Helps to advance a common language for the mixed research field
4. Provides guidance and direction for researchers to design their mixed research studies
5. Can be used to help readers and consumers of mixed research identify the type of mixed analysis used by the researcher
6. Can be used to enhance the instruction of mixed research courses (Leech & Onwuegbuzie, in press; Teddlie & Tashakkori, 2003).
   

Fundamental Principle of Mixed Analysis

When conducting a mixed analysis, researchers should consider the fundamental principle of mixed analysis (Tashakkori & Teddlie, 2003b). According to this ‘fundamental principle,’ a mixed analysis involves the use of quantitative and qualitative analytical techniques that are utilized either concurrently or sequentially, sometime after the data collection process, from which interpretations are made either in a parallel, integrated, or iterative manner. When quantitative and qualitative analytical techniques are utilized concurrently (ie concurrent mixed analysis), results stemming from one data analysis phase (eg quantitative analysis) do not inform the results stemming from the other phase (eg qualitative analysis). In a recent review study by Bryman (2006), this concurrent gathering of quantitative and of qualitative data appear to be a very common way in which mixed studies are conducted. That is, in 57% of the social science studies that were examined, quantitative data were gathered through surveys and qualitative data through interviews; and in 27% of the studies Bryman (2006) reviewed, quantitative data were gathered through surveys and qualitative data through open-ended questions on these same surveys. In neither of these groupings did the quantitative data or the qualitative data inform the results obtained in the other phase.

In mixed research studies, findings obtained from both analysis phases may be interpreted separately (ie parallel interpretation) or combined and interpreted at the data interpretation stage (ie integrated interpretation). Conversely, when quantitative and qualitative analytical techniques are utilized sequentially (ie sequential mixed analysis), the qualitative phase is conducted first, which then informs the subsequent quantitative phase, or vice versa. That is, a sequential mixed analysis involves some or all of the findings stemming from one analysis phase being utilized to drive or inform the second analysis phase. In other words, findings obtained from both analysis phases are interpreted iteratively.

Finally, a mixed researcher might undertake a parallel mixed analysis in which quantitative and qualitative data are collected either concurrently or sequentially, but then quantitative and qualitative results are interpreted separately. The most common form of parallel mixed analyses is when the mixed researcher, once both sets of analyses have been conducted and verified, interprets and writes up the two sets of findings separately. The mixed researcher might even publish the two sets of findings/interpretations in different outlets.

Mixed Analysis Matrix

Before conducting a mixed analysis, four decisions need to be made by the investigator. First, the investigator needs to determine the number of data types that will be analyzed. This decision is based on the number of data types obtained during the data collection stage of the mixed research process. Data types can be represented either by quantitative or qualitative data. Quantitative data may include responses from standardized tests, rating scales, self-reports, symptom checklists, or personality inventories. Qualitative data may include interview responses, open-ended survey responses, observations, personal journals, diaries, permanent records, transcription of meetings, social and medical histories, poems, and photographs. If only one data type (ie quantitative only or qualitative only) is used, then we refer to this as monotype data. The collection of monotype data typically occurs in mixed studies when data are collected for only one sample. Conversely, if both data types are used, we refer to this as multitype data. Multitype data are collected frequently in mixed research because the majority of mixed studies involve the collection of data from two or more samples (cf. Collins, Onwuegbuzie & Jiao, 2007).

Second, the investigator should determine the data analysis types intended to be utilized. These data analysis types either can be quantitative (ie statistical) or qualitative (eg narrative, visual). If only one data analysis type (ie quantitative analysis only or qualitative analysis only) is used, then we refer to this as mono-analysis. Conversely, if both data analysis types are used, then we refer to this as multi-analysis.

Third, the investigator should decide whether the qualitative and quantitative analyses will be carried out concurrently or sequentially. That is, will the qualitative data and the quantitative data be analyzed at approximately the same points in time during the study (ie concurrent data analysis), or will one set of data be analyzed prior to analyzing the other dataset (ie sequential data analysis)? For a recent study in which concurrent data analytic techniques were used, readers are referred to Luck, Jackson, and Usher (2007). In their study of 20 registered nurses, qualitative data were obtained from participant observations, semi-structured interviews, informal field interviews, and journaling by the researchers. Luck et al generated their quantitative data via a structured observation guide in which frequency counts were obtained. Both sets of data were created independently of each other at approximately the same time during the study. O’Cathain et al (2004)’s inquiry provides an example of a study in which a sequential analysis was utilized.

The purpose of this investigation was to examine the consistency of nurses’ triage decisions, the relationship between nurses’ clinical backgrounds and variations in decisions, and nurses’ perceptions regarding factors that impacted the clinical decision-making process. In the qualitative phase, semi-structured interviews were undertaken utilizing a sample of 24 nurses to examine their attitudes toward the decision-support software used with regard to a 24-hour telephone medical helpline, their use of the software to inform decisions concerning health care, and the degree that nurses’ clinical backgrounds influence the type of advice given to callers. The quantitative phase consisted of an analysis of software log data collected on all triaged calls during a one-month period and the clinical and demographic characteristics of 296 nurses who were employed by the helpline service during this period. Qualitative data were analyzed to identify hypotheses for the quantitative phase and to provide insights for interpreting the quantitative data.

Fourth, the investigator needs to determine whether the qualitative or quantitative analyses will be given priority, or whether they will be assigned equal status in the mixed analysis. An example of a study in which the qualitative analysis was given much more weight than the quantitative analysis is the study conducted by McVea et al (1996). In this investigation, McVea et al assessed the effectiveness of a program entitled ‘Put Prevention into Practice’ when it was utilized by family physicians in eight Midwestern private practice settings. The quantitative analysis, the phase with the least weight, involved analyzing office environment and clinical encounters checklists and chart audits. The qualitative phase involved analysis of the following data: participant observation of clinic operations and patient encounters, responses from semi-structured and key informant interviews with physicians and staff members, chart examination data, and structured post-patient encounter and office environment checklists. An example of a study in which the quantitative analysis was given more weight than the qualitative analysis is the study conducted by Bottge, Heinrichs, Mehta, and Hung (2002). These researchers described and compared the mathematical performance of students with disabilities by utilizing qualitative research strategies implemented in a quasi-experimental, non-equivalent control group design. Quantitative data consisted of students’ scores on mathematical tests designed to assess computational and word problem performance, administered as pre- and post- test measures. Scores generated from these measures were analyzed utilizing analyses of covariance. Classroom observational data and student interview data were collected to augment researchers’ interpretations of the quantitative data.

Onwuegbuzie and DaRos-Voseles’ (2001) investigation provides an example of a study in which the quantitative and qualitative analyses were approximately equally weighted. These researchers compared the effectiveness of cooperative learning (CL) versus individual learning (IL) among graduate students enrolled in a research methodology course. A total of 81 students were enrolled in the sections in which cooperative learning groups were formed to undertake the major course requirements and 112 students were enrolled in the sections wherein assignments were undertaken individually. The quantitative analysis involved comparing statistically the CL and IL groups’ performance on the midterm and final examinations. The qualitative analysis involved an in-depth examination of the both sets of students’ reflective journals about their experiences in their respective research classes. These four decisions help to generate what we refer to as a mixed analysis matrix, which is presented in Table 1.4, 5 However, the first two decisions (ie number of data types and number of data analysis types) are the backbone of the mixed analysis matrix in which the number of data types (ie quantitative or qualitative vs. quantitative and qualitative) and the number of data analysis types (ie quantitative or qualitative vs quantitative and qualitative) are crossed. This 2x2 representation yields four cells. Each of these cells is described in the following paragraphs.

Cell 1

The first cell represents analysis of one data type (eg qualitative) using one analysis type (eg qualitative). As such, this cell contains traditional monotype monoanalyses, which involve either a quantitative (ie statistical) analysis of quantitative data or a qualitative analysis of qualitative data. Such analyses indicate that the underlying study either is quantitative or qualitative in nature, respectively - neither of which represent mixed research. Although our interest in this article is in the mixed analysis approaches, the Mixed Analysis Matrix presented in Table 1 incorporates analyses involved in all three research paradigms (ie quantitative, qualitative, mixed).

Cell 2

The second cell represents analysis of two data types (ie qualitative and quantitative) using only one data analysis type - that is, multitype monoanalyses. Because only one type of analytical technique is used, the analysis is not mixed. Specifically, one type of analytical technique is used because it is generally assumed that qualitative data cannot be analyzed by using quantitative analysis techniques exclusively and that quantitative data cannot be analyzed by using qualitative analysis techniques exclusively. Therefore, this cell either involves a quantitative analysis of quantitative data or a qualitative analysis of qualitative data, but not both.

As such, one data type will not be analyzed, leading to an incomplete analysis. Thus, multitype monoanalyses should be avoided because it is unlikely that they would lead to the research question(s)—that presumably necessitated the use of mixed research approaches—being fully addressed. Multitype monoanalyses tend to occur when the data analyst only is able to perform one type of analysis (ie quantitative analysis only or qualitative analysis only). Indeed, this situation may occur in the context of an analyst conceding, for example, that ‘In addition to the quantitative data, I have plenty of qualitative data but I do not know what to do with this qualitative data.’ Thus, we suggest that to avoid multitype monoanalyses, a team of analysts, versed in quantitative and qualitative analytical techniques should be used who, between them, are able to analyze both the qualitative and quantitative data.

Cell 3

The third cell represents analysis of one data type (eg quantitative) using both data analysis types (ie qualitative and quantitative). This class of analysis is called a monotype mixed analysis. Because both quantitative and qualitative analytical techniques are used, the analysis is mixed. The first analysis employed in this cell should directly match the data type. Thus, if the data type is quantitative, then the first phase of the mixed analysis should be quantitative (ie statistical). Similarly, if the data type is qualitative, then the first phase of the mixed analysis should be qualitative. Data that stem from the initial analysis then are converted into the other data type. That is, the quantitative data are transformed into data that can be analyzed qualitatively or what is known as qualitizing data (Tashakkori & Teddlie, 1998), or the qualitative data are transformed into numerical codes that can be analyzed statistically or what is known as quantitizing data (Tashakkori & Teddlie, 1998).

As of this time, relative few studies are available in which data have been transformed, either through qualitizing or through quantitizing data. Greene et al (1989) reported that only 5 of 57 evaluation studies involved transformed data, whereas Bryman (2006), much more recently, noted that only 7 of 232 studies transformed data in their analyses. Readers are referred to a recent study by Wang, Gibson, Salinas, Solis, and Slate (2007) for an example of a study involving qualitative data that were transformed into quantitative data for statistical analyses. In their study, the mission statements of 102 2-year and 4-year colleges were analyzed to determine the extent to which their mission statements were similar in nature. Fifteen themes were determined to be present across their sample. Once themes had been identified, they were then quantitized or transformed into quantitative information. When a theme was identified as being present in a particular college’s mission statement, then a value of 1 was assigned. When a theme was identified as not being present, then a value of 0 was assigned. Thus, each of the 102 colleges had a series of 1s and 0s, depending on whether the identified themes were present in their mission statements. This data transformation, from qualitative themes transformed to provide quantitative information, permitted Wang et al to use chi-square analyses and discriminant analyses. As computer software, especially qualitatively based software, becomes more sophisticated, the trend toward more researchers engaging in transforming data would be probable (Bazeley, 2006).

Qualitizing data

One way of qualitizing data is via narrative profile formation (ie modal profiles, average profiles, holistic profiles, comparative profiles, normative profiles), wherein narrative descriptions are constructed from statistical data. For example, Teddlie and Stringfield (1993) conducted a longitudinal study of eight matched pairs of schools initially classified either as effective or ineffective with regard to baseline data. Five years after the study was initiated, these researchers used eight empirical criteria to re-classify the schools’ effectiveness status. These criteria were:

1. Norm-referenced test scores
2. Criterion-referenced test scores
3. Time-on-task in classrooms
4. Scores on quality of classroom instruction measures
5. Faculty stability
6. Student attendance
7. Changes in socioeconomic status of the schools’ student bodies, and
   
8. Measures of school climate.
   

Teddlie and Stringfield converted these quantitative data (ie qualitized them) into the following four qualitatively-defined school profiles: (a) stable more effective, (b) stable less effective, (c) improving, and (d) declining. These school profiles then were used to add greater understanding to the researchers’ longitudinal, evolving perspectives on the schools’ effectiveness status.

Quantitizing data

When researchers quantitize data, ‘qualitative ‘themes’ are numerically represented, in scores, scales, or clusters, in order more fully to describe and/or interpret a target phenomenon’ (Sandelowski, 2001, p.231). Quantitizing often involves reporting effect sizes associated with qualitative observations (Onwuegbuzie, 2003; Sandelowski & Barroso, 2003), which can range from manifest effect sizes (ie counting qualitative data to determine the prevalence rates of codes, observations, words, or themes) to latent effect sizes (ie quantifying non-observable content, for example, by factor-analyzing emergent themes; cf. Onwuegbuzie 2003). For instance, Sandelowski and Barroso (2003) showed how effect sizes can be utilized to conduct metasummaries of qualitative findings. These researchers defined a qualitative metasummary as ‘a form of systematic review or integration of qualitative findings in a target domain that are themselves topical or thematic summaries or surveys of data’ (p.227). They conducted a qualitative metasummary of 45 published and unpublished reports of qualitative studies of HIV-positive women with findings on motherhood. This metasummary led to 800 findings being extracted, which were reduced to 93 abstracted findings, from which manifest effect sizes were calculated.

Sandelowski and Barroso found that five results had effect sizes ranging from 25% to 60%, with both published and unpublished articles contributing approximately equally to the strength of these findings. A total of 73 findings had effect sizes that were less than 9%, with 47 of them having effect sizes of only 2%. (For an extensive discussion of effect sizes in qualitative research, see Onwuegbuzie, 2003; Onwuegbuzie & Teddlie, 2003; and Sandelowski & Barroso, 2003.) In Cell 3, the investigator also has to decide whether the qualitative or quantitative analyses will be given priority, or whether they will be given equal status in the mixed analysis. This decision, which stems from the goals and objectives of the mixed research as well as from the rationale for mixing, is important because it allows the investigator to determine the level of sophistication of one analysis relative to the other analysis. Readers are referred to Maeda (2006) for an equal-status mixed research study. In this study of 80 Brazilian immigrants, Maeda (2006) quantitatively analyzed responses to surveys completed by the study participants and qualitatively analyzed transcripts generated from face-to-face interviews with 8 Brazilian sample members.

For a dominant-less dominant example, Mactavish and Schleien (2000) conducted a mixed research study on family recreation in which the qualitative analysis of interview data was dominant and the quantitative analysis of surveys was less dominant. Another dominant-less dominant example, this time where the quantitative analysis represented the dominant analytical phase and the qualitative analysis played a lesser role, is provided in a study by Boatwright and Slate (2000). In this study, the authors used qualitative analyses from interviews, focus groups, and documents to create a quantitative measure of work ethics which they then administered to 307 persons.

Cell 4

The fourth cell represents analysis of both data types (eg quantitative and qualitative) using both analysis types (ie qualitative and quantitative). This class of analysis is called a multitype mixed analysis. Because both quantitative and qualitative analytical techniques are used, the analysis is mixed. The multitype mixed analysis might be concurrent involving a statistical analysis of the quantitative data combined with a qualitative analysis of the qualitative data, followed by meta-inferences being made in which interpretations stemming from the quantitative and qualitative findings are integrated some way into a coherent whole (Tashakkori & Teddlie, 2003b).

In both the multitype concurrent and multitype sequential mixed analyses, either the quantitative or qualitative analyses can be dominant or these analyses can share equal status. An example of a concurrent analysis of quantitative data combined with qualitative analyses leading to a meta-inferences being formulated is a study conducted by Mactavish and Schleien (2004). These researchers examined the role of family recreation in families that included a child with a developmental disability. Survey responses (ie quantitative data) and interview responses (ie qualitative data) were collected concurrently. The survey responses were analyzed utilizing descriptive and non-parametric statistical techniques and themes derived from the interview data were developed using constant comparative methods. Subsequently, both data types were interpreted at the data interpretation stage. Alternatively, the multitype mixed analysis could be sequential in nature in which findings from the qualitative analysis inform the subsequent quantitative analysis, or vice versa.

An example of a sequential analysis of qualitative data combined with quantitative data leading to meta-inferences being formulated is a study conducted by Onwuegbuzie, Witcher, Collins, Filer, Weidmaier, and Moore (2007). These researchers conducted a sequential qualitative-quantitative mixed analysis to examine college students’ perceptions of characteristics of effective college teachers. The first stage analysis comprised a 5-step qualitative analysis of students’ responses to an open-ended survey. This analysis produced nine themes. Next, the nine themes were transformed (ie quantitized) to provide descriptive data (ie frequency data). These frequencies were interpreted to determine a prevalence rate for each theme. The third stage of the analysis consisted of an exploratory factor analysis to determine meta-themes (ie clusters of themes) based upon the nine themes’ underlying structure. This analysis generated four meta-themes.These three stages of analysis led to the development of what the researchers called the ‘CARE-RESPECTED Model of Teaching Evaluation’ (p.151) because it represented the nine themes and the four meta-themes, the names of which were arranged to produce the acronyms CARE (ie Communicator, Advocate, Responsible, Empowering) and RESPECTED (ie Responsive, Enthusiast, Student-Centered, Professional, Expert, Connector, Transmitter, Ethical, Director).

The fourth stage of the analysis was conducted to examine the relationship between the nine themes and the sample’s demographic variables. Finally, the study’s findings were interpreted in terms of the score validity of college teaching evaluation forms, in general. Although not represented in the Mixed Analysis Matrix, Cell 4 can accommodate much more complex analytical designs. For example, any of the nine analytic designs in Cell 4 could include quantitizing and/or qualitizing data. Also, both Cell 3 and Cell 4 can be expanded to include iterative data analysis techniques. An example of an iterative data analysis is what Li et al (2000) termed cross-tracks analysis. This analysis is characterized by a concurrent analysis of both qualitative and quantitative data such that the data analysis oscillates continually between both sets of data types throughout various stages of the data analysis process (Li et al 2000). However, the six mixed analytic designs in Cell 3 and the nine analytic designs in Cell 4 provide the researcher with many choices.

Mixed Analysis Steps

Onwuegbuzie and Teddlie (2003) have identified the following seven stages of the mixed analysis process:

1. Data reduction
2. Data display
3. Data transformation
4. Data correlation
5. Data consolidation
6. Data comparison, and
   
7. Data integration.
   

These authors defined data reduction as reducing the dimensionality of the quantitative data (eg via descriptive statistics, exploratory factor analysis) and the qualitative data (eg via interim analysis, thematic analysis, memoing). This step is pertinent to analyses in all four cells of Table 1. The next step, data display, refers to describing visually the quantitative data (eg tables, graphs) and/or qualitative data (eg graphs, charts, matrices, checklists, rubrics, networks, and Venn diagrams). This step also is pertinent to analyses in all four cells of Table 1.

This step is followed, if needed, by the data transformation stage, in which data are quantitized and/or qualitized. This step is pertinent only to Cells 3 and 4 of Table 1. Data correlation, the next step, involves qualitative data being correlated with quantitized data or quantitative data being correlated with qualitized data. This step is relevant only for Cell 4 because both types of data are needed for a correlational analysis. This step is followed by data consolidation, wherein both quantitative and qualitative data are combined to create new or consolidated codes, variables, or data sets. Again, this step is relevant only for Cell 4. The next stage, data comparison, involves comparing the findings from the qualitative and quantitative data sources or analyses - again representing only Cell 4. Data integration is the final stage of the mixed analysis process, wherein both qualitative and quantitative findings are integrated into either a coherent whole or two separate sets (ie qualitative and quantitative) of coherent wholes - another feature of Cell 4.

Table 2 illustrates the stages of the mixed analysis process that can occur for each of the four cells of the Mixed Analysis Matrix. In Cell 1 (ie Monotype Monoanalysis), which involves either a quantitative analysis of quantitative data or a qualitative analysis of qualitative data, only the data reduction and data display stages are possible because only one type of data are collected and the only analysis that occurs is a traditional one—in which both the type of data and analysis can be classified as representing the same paradigmatic tradition (ie either quantitative or qualitative) - what we refer to as a within-paradigm analysis. Thus, the sole data type collected is not converted (ie data transformation); nor is data correlation, data consolidation, data comparison, or data integration possible.

Similarly, in Cell 2 (ie Multitype Monoanalysis), because only one type of data is analyzed using a within-paradigm analysis, only the data reduction and data display stages are applicable. In Cell 3 (ie Monotype Mixed Analysis), in addition to the data reduction and data display stages, data transformation can occur. In fact, data transformation is a hallmark of Cell 3 because there is only one type of data collected, which necessitates what we refer to as between-paradigm analysis - which involves using an analysis technique that is more associated with one traditional paradigm (eg quantitative) to analyze data that originally represented the type of data collected that are more often associated with the other traditional paradigm (eg qualitative). Thus, some form of transformation (ie quantitizing, qualitizing) is needed for the analysis technique to be compatible with the data type.

In stark contrast to the other three cells, in Cell 4 (ie Multitype Mixed Analysis), all seven stages of the mixed analysis process are possible because it involves both data types and both classes of analysis (ie quantitative and qualitative). Thus, depending on the research goal(s), research objective(s), rationale and purpose for mixing, and research question(s) underlying the mixed study, the analyst has the option to utilize both within-paradigm analyses and between-paradigm analyses. As such, Cell 4 has the great potential to yield the most in-depth and rigorous analysis possible, relative to the analysis conducted in monomethod studies.

Rationale and Purpose for Conducting Mixed Analyses

According to Collins, Onwuegbuzie, and Sutton (2006), significance enhancement is one of the four major rationales for mixing quantitative and qualitative approaches. This rationale type is particularly relevant when designing and conducting mixed analyses because it refers to enhancing researchers’ interpretations of data. Significance enhancement includes Greene et al’s (1989) five purposes for conducting mixing approaches, namely, tri­angulation (ie compare findings from the quantitative data with the qualitative results), complementarity (ie seek elaboration, illustration, enhancement, and clarification of the results from one method with findings from the other method), development (ie use the results from one method to help inform the other method), initiation (ie discover paradoxes and contradictions that culminate in a re-framing of the research question), and expansion (ie expand breadth and range of a study by using multiple methods for different study phases).

Decisions made regarding these five purposes help determine which type of mixed analysis should be undertaken: a parallel mixed analysis, concurrent mixed analysis, or sequential mixed analysis. Specifically, as can be seen in Table 3, if the purpose of mixing is complementarity, then all three families of mixed analyses (ie parallel, concurrent, and sequential) can be utilized. If triangulation or initiation is the purpose, then parallel and concurrent mixed analyses are appropriate. If development is the purpose, then concurrent and sequential mixed analyses are pertinent. Finally, a purpose of expansion lends itself to a sequential mixed analysis.

Table 4 illustrates the significance enhancement rationale types (ie tri­angulation, complementarity, development, initiation, and expansion) that can occur for each of the four cells of the Mixed Analysis Matrix. Because Cell 1 (ie Monotype Monoanalysis) and Cell 2 (ie Multitype Monoanalysis) do not involve any form of mixed analysis, none of these rationale types are pertinent. In Cell 3 (ie Monotype Mixed Analysis), the rationale for enhancing the significance of findings might be complementarity, expansion, and/or development. In Cell 4 (ie Multitype Mixed Analysis), all five rationale types are pertinent - again highlighting the potential flexibility of mixed analyses that fall under this category.

Summary and conclusions

The purpose of this article was to provide a typology of mixed-analysis techniques to assist researchers in analyzing data from mixed research studies - specifically, to outline the available analysis techniques for multiple types of data. We discussed the fundamental principle of mixed analysis, presented the Mixed Analysis Matrix, described the steps in a mixed analysis, and delineated the rationale and purpose for conducting mixed analyses.

Outlining the mixed-analysis techniques available for researchers hopefully will increase awareness of the number of choices for analyzing data obtained from mixed research studies. Being aware of the type of data that has been collected, delineating the type of mixed analysis, identifying the purpose for conducting mixed analysis, and demarcating the significance enhancement rationale type and the relationship between these, can increase the depth of the analysis and increase the rigor of the study.

Toggle references

Navigate: Archives | Contents | Top of page

Search

Related Publications from eContent Management