• No results found

Data analysis methods / approaches

Chapter 3 – Research methodology

3.4 Data analysis methods / approaches

3.4.1 Data analysis process

Qualitative research is creative and interpretive and so is the data analysis (Denzin & Lincoln, 1998) which often needs to be specially designed for particular research (Maxwell, 2005). It is possible to approach data analysis in multiple ways since “there is

no single interpretive truth” (ibid., p. 30) and material can be interpreted in many ways (Miles & Huberman, 1994).

Data analysis can be regarded as an interactive and cyclical process consisting of three interrelated activities – data reduction, data display and the drawing of conclusions and verification (see Figure 3.3) (Miles & Huberman, 1994). Data reduction deals with sharpening, sorting, focusing, discarding and organizing data in order to facilitate the drawing of conclusions (ibid.) Data reduction takes place throughout the research process. Data display refers to “an organized, compressed assembly of information that permits conclusion drawing and action” (ibid., p. 11). Data display may take the form of charts, graphs, networks and matrices. Conclusion drawing and verification involves deciding on the meaning of data and the further testing of these meanings “for their plausibility, their sturdiness, their “confirmability” – that is, their validity” (ibid., p. 11; original emphasis). The examples of the interactive and cyclical process in data analysis can be seen very often in qualitative research such as incorporating patterns that emerge from data already collected into further interviews by using the ‘constant comparison’ method (Jordan & Gibson, 2004).

Figure 3.3 Components of Data analysis: Interactive Model (Source: Miles & Huberman, 1994, p. 12)

Data collection Data reduction Data display Conclusions: drawing / verifying

The aim of data analysis is to divide or break down information into constituent parts by dissecting, reducing, sorting and reconstituting data (Spiggle, 1994). The process of data analysis includes such operations as categorization or coding, abstraction (grouping categories into higher-order conceptual constructs), comparison (noting similarities and differences within collected data that informs further data collection), dimensionalization (identifying properties and categories of constructs), integration (building theory by integrating categories and constructs), iteration (allowing preceding data collection and analysis operations to shape subsequent ones) and refutation (subjecting emerging inferences to empirical scrutiny) (ibid.).

As stated earlier, there is no particular set of methods associated with doing qualitative research; the same is true for analysing qualitative data. Approaches such as semiotics, narrative, content, discourse, archival and phonemic analysis and even statistics can be used in qualitative research (Denzin & Lincoln, 1998). The particular analysis technique will also depend on the type of data being analysed, i.e. whether the qualitative data appears in the form of words or numbers, or in still or moving images (Miles & Huberman, 1994).

Most analysis however is done with text which can be regarded either as an object of analysis or as a proxy for experience (Denzin & Lincoln, 2000). If text is regarded as an object of analysis, then conversation, performance and narrative analysis as well as analysis of grammatical structures can be undertaken (ibid.). If text is treated as a proxy for experience, then either a systematic elicitation approach or analysis of free-floating text can be used (ibid.). In the case of the latter, analysis can be done based on words (e.g. word count, semantic networks and cognitive maps) or codes (e.g. grounded theory, content analysis, schema analysis) (ibid.).

3.4.2 Data analysis approach in this study

The preliminary data analysis began during the interviewing stage by reflecting on already-conducted interviews which helped to re-design questions for further interviews (Rubin & Rubin, 1995). The following steps in data analysis included writing down notes

while transcribing and later reading through interviews which assisted in developing tentative ideas and facilitated analytical thinking (Maxwell, 2005).

Once the interviews had been transcribed, data was analysed further. It should be noted that several attempts were made to approach the data analysis before an appropriate approach that suited this study was found. This included coding the interview transcripts manually, setting up constructs tables in Microsoft Word where relevant excerpts from interviews were copied, and developing initial codes that were recorded in the codebook. After initial attempts to code the data, the researcher developed the following approach to data analysis.

Firstly, data was categorized according to organizational categories that served as “bins” for sorting interview data for further analysis (Maxwell, 2005). This allowed qualitative data to be organized around key issues (Patton, 2002). This process is sometimes also called ‘theming of data’ (Saldaña, 2009). Constructs identified through research objectives and questions (see Table 2.1) formed the basis for organizational categories, for example, “Slow tourist concept” or “Slow philosophy in destination marketing.” Because of the large amount of data collected, NVivo10 software was used to assist the data analysis (Bazeley & Jackson, 2013). Interview transcripts were imported in NVivo10 and the data was coded or grouped in nodes according to the constructs. It is important to emphasize that NVivo10 was used as a data organization and management tool and that the actual analysis was performed by the researcher.

Secondly, the nodes or constructs were exported to separate Word documents then printed out and analysed by manually coding on paper. The resulting codes were then put into separately created Word documents which facilitated a process of overseeing and subsequent further analysis. This process helped to further refine the codes and identify emerging patterns.

It should be noted that the analysis of data was undertaken independently, in consultation with supervisors.

3.4.3 Coding, thematic analysis, and making sense of data

A thematic analysis was used to analyse the data (Boyatzis, 1998). A theme can be seen as “an outcome of coding, categorization, an analytic reflection” (Saldaña, 2009, p. 13; original emphasis). Thematic analysis is different from content analysis which a “systematic, objective, quantitative analysis of message characteristics” (Neuendorf, 2002, p. 1). As such, themes are derived from codes.

As mentioned earlier, coding or categorization is a data analysis operation and codes are used to analyse free-floating text (Denzin & Lincoln, 2000). Coding is regarded as a “transitional process between data collection and more extensive data analysis” that helps a researcher to organize and group data into categories that share similar characteristics forming a pattern (Saldaña, 2009, p. 4).

A code can be “a word or short phrase that symbolically assigns a summative, salient, essence-capturing and / or evocative attribute for a portion of language-based or visual data” (Saldaña, 2009, p. 3). The process of coding however is more of “an interpretive act” rather than “precise science” (ibid., p. 4) and both inductive and deductive coding is possible (Spiggle, 1994).

Codes can be developed in three ways. They can be driven by 1) theory, 2) prior data or 3) raw data (Boyatzis, 1998). This study employed a combination of theory-driven and data-driven approaches. Existing models and frameworks have been used to define somea prioricodes to build an initial coding template. However, the codes were revised as analysis went on. New codes emerged as the researcher constructed the codes inductively from the raw data. Substantive categories were developed using descriptions that research participants provided when talking about constructs (Maxwell, 2005).

Several coding methods and cycles were used in this study. The combination of structural, descriptive and In Vivo coding was used in the first cycle coding (Saldaña, 2009). Following that, pattern coding was used during the second cycle of coding in order to identify emerging themes (Miles & Huberman, 1994). Theoretical categories were developed by placing coded data in a more general framework (Maxwell, 2005).

During the data analysis process, thick descriptions were provided. It is important to note that thick descriptions are not merely lengthy descriptions but rather an outcome of cultural analysis, i.e. interpretive analysis of cultural meanings (Wallendorf & Brucks, 1993). Creating a rich, detailed and concrete description forms a basis for data interpretation (Patton, 2002).

The final step involved making sense of the data through interpretation. Data interpretation deals with making sense of the data or grasping its meaning through more abstract conceptualizations (Spiggle, 1994). In interpretation, there are no guidelines to follow like in data analysis since “interpretation occurs as a gestalt shift” and springs from serendipity and mental activities (ibid., p. 497). There are several tactics for generating meanings, such as noting patterns and themes, seeing plausibility, clustering, making metaphors, counting and making contrasts and comparisons (Miles, Huberman, & Saldaña, 2014). One of the most commonly used strategies for grasping the meanings and experiences of informants is seeking coherent patterns in meanings (Spiggle, 1994).

This study employed an active search for patterns and themes in the data. The researcher noted recurring patterns that brought together separate pieces of data (Miles et al., 2014). The data analysis and interpretation was then organized around these key themes. The results of the data analysis and interpretation are presented in Chapter 4.