Chapter 3 : Methodology
3.9 Data Analysis and Interpretation
The power of multi-method research in integrating separate data points from multiple methods has been confirmed in the research literature (Brewer & Hunter 1989). There are numerous recommendations within the literature for multi-method researchers to combine results, rather than reporting them in separate sets (McMurray, Pace & Scott 2004). According to Cresswell and Plano Clark (2007, P.83) “A study that includes both quantitative and qualitative methods without explicitly mixing the data derived from each is simply a collection of multiple methods”.
As Kemper et al. (2003, P.284) reiterate, multi-methods research generally empowers the researcher for the triangulation of data sources, and forms a multifaceted view of the research questions. Discussing the practical problems associated with triangulation, Flick (2004) distinguished between case triangulation and the triangulation of data sets. Case triangulation, as Flick observed, is the application of triangulated methods to the study of the same cases, which will can overload on participants and as a consequence, increase the danger of dropout. The Triangulation of data sets however, occurs when the methods are implemented independently. Triangulation happens later at the data analysis stage, in order to assess the data emerging in separate sets in relation to each other. The practical problem here, as Flick (2004, P.182) points out, is “How comparability of the samples, where the different methods have been applied, can be guaranteed”. This was indeed, an issue with the present research because, as discussed earlier, this research includes the implementation of between-methods triangulation on two distinct sets of samples, what Flick calls the triangulation of data sets. However, as notified elsewhere “Different methods might simply be tapping different dimensions, qualities, or aspects of a given phenomenon” (Hunter & Brewer 2003, P.581).
Onwuegbuzie and Teddlie (2003) considered two major rationales for mixed-method data analysis, namely those of representation and legitimation. Representation as they argue, is the generation of more meaning by extracting more, adequate information out of the data. Legitimation however, is an attempt to assess and document the rigor or legitimacy of the work, including the validity, trustworthiness, transferability, etc. of the interpretations. These two rationales seem to be compatible with the goals of multi-method studies in general, as noted earlier in this chapter. The main objective of the research therefore, as previously mentioned, was not to compare the results, but to elucidate divergent perspectives. Another problem that has been mentioned in relation to this type of study, is the possible influence of time difference, emerging from the sequential sequencing of the research problem (Flick 2004). This, however, was unlikely to affect the current research, because the time interval between the two phases of the research was just a few months. Accordingly, along with McMurray et al.’s suggestion, and to allow a more inclusive understanding of the topic, a convergent analysis of data from both the survey and case-studies was conducted in the present research.
As maintained widely throughout the literature, decisions about data analysis techniques in mixed-methods research should be based on the purpose of the research (Onwuegbuzie & Teddlie 2003). Onwuegbuzie and Teddlie (2003, P.352) define mixed-method data analysis as “The use of quantitative and qualitative analytical techniques, either concurrently or sequentially, at some stage beginning with the data collection process, from which interpretations are made in either a parallel, an integrated, or an iterative manner”.
Discussing data analysis within mixed-methods designs (See Figure 3.3) Cresswell and Plano Clark (2007) also refer to the existence of a concurrent form of analysis for triangulation design, and provide researchers with a general guideline for its implementation. This involves the conduct of a separate initial analysis of quantitative and qualitative datasets at the first stage, and merging the two datasets in the later stage to develop a complete picture of the topic. According to them, there are two different techniques for merging different types of data. The first technique includes transforming or converting one set of data into the other form, what elsewhere would have been termed quantitizing qualitative data or qualitizing quantitative data, to
make the two types of datasets comparable (Tashakkori & Teddlie 1998, P.125). In choosing this technique therefore, the researcher had the option of choosing to employ one of the two following analytical strategies including;
• Concurrent analysis of the same data from two methods through qualitizing
• Concurrent analysis of the same data from two methods through quantitizing (Tashakkori & Teddlie 1998, P.128).
The second technique as Cresswell and Plano Clark (2007) also reveal, is to compare the data without conversion through either a matrix or through discussion. The analytical strategy associated with this technique has been known as parallel mixed analysis (Tashakkori & Teddlie 1998). Concurrent (simultaneous) analysis of both quantitative and qualitative data also known as triangulation of data sources has been described by Tashakkori and Teddlie (1998) as the most widely used analytical strategy in social and behavioural mixed-method studies. Parallel mixed analyses as Onwuegbuzie and Teddlie (2003, P.366) confirm, may be used if the purpose of the mixed-method research is triangulation, complementarity, initiation, or expansion.
The comparison of datasets through discussion is the technique used in the current research. This analytical technique has been noted by Cresswell and Plano Clark as one frequently used in parallel mixed analysis, which occurs when the researcher examines the similarities of the results from both datasets in the discussion section of the study. In this approach, the researcher first presents a statistical result (in the form of descriptive or inferential statistics) from the quantitative part of the research, following it up with relevant quotes or information about a theme from the qualitative results, to either confirm or disconfirm the first mentioned quantitative results or vice versa.
Figure 3.3 Concurrent mixed data analysis procedures and strategies in triangulation design.
An Adaptation of Cresswell and Plano Clark (2007, P.137) .
Stage 1 Initial data analysis
Analysis of quantitative data Analysis of
qualitative data
Stage 2
Merging the two data sets
Data transformation
Comparison of the results
Matrix Discussion Quantitizing Qualitizing Concurrent analysis of the same data Parallel mixed analysis
3.10 Chapter Summary
This chapter has reported how the researcher conducted the research process and executed data collection and analysis. Overall, the study was not designed to test any specific hypothesis or existing theory. The research is exploratory in nature, seeking to explore the phenomenon of KM education within the LIS sector. It falls largely within an interpretivist paradigm in that it seeks not to identify or test variables, but rather to draw meaning from social contexts, in this case from the perceptions of key players within the LIS education sector. In the epistemology of this paradigm, knowledge is seen to be derived from everyday concepts and meanings, in this case from current trends and practices in KM education. As little is known about the phenomenon under investigation, the research strategy or logic of the work is inductive. The method employed in this project comprises a combination of quantitative and qualitative approaches. Data collection was conducted through literature review and document analysis, followed by the conduct of a survey and in- depth interviews.
An integrated analysis of findings from both the questionnaire and the interviews is presented in the following chapter.
4.1 Introduction
As discussed in the methodology section, the current research has employed the complementarity model of triangulation, and has sought for complementary results (Erzberger & Kelle 2003) . Patton’s (1980, P.330) point is worth noting about there being no magic in triangulation. According to Denzin (1989, P.246), “The researcher using different methods should not expect findings generated by different methods to fall into a coherent picture. They will not and they can not, for each method yields a different picture and slice of reality. What is critical is that different pictures be allowed to emerge. Methodological triangulation allows this to happen”.
Hence, the fundamental intent of the research was to generate a fuller picture of the domain under investigation, and provide a more comprehensive, understanding of the topic. The confirmation of results has not necessarily been an objective of this research, although in certain cases this has been achieved. What is presented in the following chapters, therefore, is an attempt to shed light on the implications of KM for LIS education, and to help provide a better understanding of the issue.
In line with Cresswell and Plano Clark’s (2007) guideline on the concurrent form of analysis for triangulation design, discussed in the previous chapter, a descriptive analysis of quantitative data from the questionnaire was first conducted using SPSS 14.0 software. Then qualitative analysis of data from interviews and open- ended sections of the questionnaire was conducted, using NVivo software. To handle data management the researcher first went through interview transcriptions, identified the main themes appeared in the data and linked the chunk of data that represented a given idea or concept to its relevant theme. The data assigned to each theme then categorised into sub- themes. Relevant data from the open- ended sections of the questionnaire were also transferred into these emerging themes/sub- themes. The Parallel mixed analysis of data through discussion as discussed by the aforementioned authors was then employed as an analytical tool to merge qualitative and quantitative findings.