• No results found

3.1 RESEARCH METHODOLOGY

3.3.4 DATA ANALYSES IN MIXED METHODS

Teddlie and Tashakkori (2009) hold the view that the basic strategies for analysing qualitative and quantitative data must first be understood before one looks at its usage in mixed methods. Let us first consider them before we continue with mixed methods data analysis.

3.3.4.1 QUALITATIVE DATA ANALYSIS

During qualitative data analysis, various forms of narrative data are converted from raw material into partly processed data which are than subjected to a particular analysis scheme. It is regarded as predominantly inductive in nature, is iterative ( involving a back and forth process between data collection and analysis), and eclectic (employing an eclectic mix of the available analytical tools that best fits the data set under consideration) (Teddlie and Tashakkori ,2009).

The three general types of qualitative data analysis is firstly, categorical strategies (that break down narrative data and rearrange those data to produce categories that facilitate comparisons leading to a better understanding of the research question); secondly contextualizing

strategies (that interpret narrative data in the context of a coherent whole text that includes

interconnections among statements and events); and qualitative data displays (that are visual presentations of the themes that emerge from qualitative data analysis.

3.3.4.2 QUANTITAIVE DATA ANALYSIS

During quantitative data analysis, numeric data are analysed, using a variety of statistical techniques (ibid, 2009:256). Three different quantitative data analysis techniques are descriptive versus inferential statistics, univariate versus multivariate statistics and parametric versus non-parametric statistics.

The above authors further define descriptive methods as the procedures used for summarizing data, with the intention of discovering trends and patterns, and summarizing results for ease of understanding and communication. Inferential techniques are generated after descriptive results have been examined, normally used for testing hypotheses or for confirming or disconfirming the results obtained from the descriptive results. Univariate statistics involve

       

66

linking one variable that is the focal point of the analysis with one or more other variables while multivariate statistics link two or more sets of variables to each other (ibid, 2009). 3.3.4.3 MIXED METHODS DATA ANALYSIS

Teddlie and Tashakkori (2009) describe mixed methods data analysis as the processes whereby qualitative and quantitative data analysis strategies are combined, connected or integrated in research studies. The typology of mixed methods data analysis is organised around the five types of mixed methods design implementation processes which is parallel mixed methods data analysis, conversion mixed methods data analysis, sequential mixed methods data analysis, multilevel mixed data analysis and fully integrated mixed methods analysis.

Parallel mixed methods data analysis is the most used mixed methods data analysis strategy in the human sciences and has been associated with other design concepts such as triangulation and convergence (Teddlie and Tashakkori , 2009). They state that “it involves two separate processes, quantitative analysis of data, using descriptive/inferential statistics for the appropriate variables, and qualitative analysis of data, using thematic analysis related to the relevant narrative data” (ibid, 2009:266).

The two sets of data analysis are different but both provide an understanding of the phenomenon under study which is linked, combined or integrated into meta-inferences. When the two sets of parallel analysis is allowed to “talk to each other” it is known as “cross over tracks analysis” whereby findings from the two methodological strands are intertwined and inform one another throughout the study (ibid, 2009:269).

Conversion mixed methods data analysis takes place, when collected qualitative data are converted or transformed into numbers (quantitizing), or quantitative data are converted into narratives or other types of qualitative (qualitizing), according to Teddlie and Tashakkori (2009). They describe quantitizing narrative data as the process whereby qualitative data are transformed into numerical data that can be analyzed statistically.

Qualitizing numeric data is described as the process whereby quantitative data are transformed into qualitative categories or narrative form (ibid, 2009). Both quantitizing and qualitizing involves one data source and its conversion to the other form but inherently mixed data analysis implies that qualitative and quantitative information is used as a data source to interlink questions (ibid, 2009).

       

67

Sequential mixed methods data analysis takes place when the qualitative and quantitative strands of a study occur in chronological order in order for the analysis in one strand to emerge from or depends on the previous strand (ibid, 2009). During a sequential QUAL to QUAN analysis with typology development, a qualitative phase occurs first, followed by a quantitative phase and the analysis from the two phases are interlinked. During a sequential QUAN to QUAL analysis with typology development, a quantitative phase occurs first, followed by a qualitative phase, and the analyses from the two phases are related to one another. In an iterative sequential mixed analysis, the analyses of data from a sequential study have more than two phases.

Multilevel mixed data analysis is described as a general analytic strategy in which qualitative and quantitative techniques are used at different levels of aggregation within a study to answer interrelated research questions (ibid, 2009). It occurs when one type of analysis (QUAL) is used at one level (students) and another type of analysis (QUAN) is used at another level (classroom).

Fully integrated mixed methods analysis takes place when there is an interactive mixing of qualitative and quantitative analyses characterized as iterative, reciprocal and interdependent. This form of analysis helps to break down the barriers between the traditional qualitative thematic and quantitative statistical dichotomy in analysis (ibid, 2009).

Before we discuss the inference process on mixed methods research we must keep in mind that the assessment of quality is a process that was already followed in the other methods. Bryman, Becker and Sempik (2008) already established in research amongst social researchers that there are variations in the levels of support for various quality criteria in social science.

Findings suggest that there is support for the relevance of validity, reliability and generalizability as criteria for judging the quality of quantitative research (Bryman, Becker and Sempik, 2008). The criteria of replicability has received less support in that type of research. Respondents regarded the relevance of validity to qualitative research considerably higher than generalizability and replicability. Bryman, Becker and Sempik (2008:274) assert that “findings relating to both generalizability and transferability imply that issues to do with the ability to generalize to populations or settings are not major concern among social policy researchers”.

Their findings show a preference for a combination of traditional quantitative research criteria and qualitative research criteria for the judging of quality in mixed methods research,

       

68

and using different criteria for the quantitative and qualitative components of a mixed methods investigation (ibid, 2008). The criteria which stands out in a mixed methods research is that it should be relevant to a research question, procedures used in the research should be transparent, the findings should be integrated and the rationale for using a mixed methods approach should be outlined.

Alicia O‟Cathain (2010) provides a comprehensive framework in her assessment of quality in mixed methods research. She highlights the need for a comprehensive framework which includes offering a structured description of a complex issue with the purpose of facilitating understanding and also addressing the need of a variety of stakeholders that want to assess the quality of mixed methods research (ibid, 2010).

After evaluating the three approaches (generic, individual and mixed) to assess the quality of a mixed study, O‟Cathain (2010) expresses preference for a mixed methods approach, hereby “inferences are drawn from the whole mixed methods study-meta-inferences-not simply from each component” (ibid, 2010:535). O‟Cathain (2010) places the comprehensiveness of the mixed methods approach at the core of her framework which consists of certain domains which is planning quality, design quality, data quality, interpretive rigor, inference transferability, reporting quality, synthesizability and utility.

The next section focuses on Teddlie and Tashakkori‟s inference process and their proposed integrative framework to assess quality.