CAPÍTULO II: PRESENTACIÓN DEL SISTEMA DE ACTIVIDADES PARA
2.3. Validación de la propuesta
As noted in Section 4.4, this study uses a qualitative approach to data collection partially driven by the desire to meet the demands of current research, which calls for studies that explore engagement from a perspective that seeks in-depth understanding rather than simple explanation. In light of this, in this section I present thematic analysis as the analytical approach adopted in this investigation to achieve such deep understanding, which can be specifically understood as uncovering the potential of the ProE model (if any) for student ASEC engagement.
Although it may seem obvious, it must be stressed that the thematic approach of data analysis only applies to the qualitative data gathered in the study and therefore omits the data obtained in the after-project questionnaires. As previously explained (Section 4.4.4), such questionnaires were only used to complement the qualitative data, as well as to develop other qualitative instruments; for this reason they are not analysed using formal or systematic procedures. In addition, it is also noteworthy that the thematic approach to data analysis only applies to the analysis conducted as a researcher (rather than as a teacher). This is because, as we will see in Chapters Six and Seven, in each research cycle I conduct two separate analyses of the data. The first one is asystematic (i.e., repeated reading of notes and questionnaire results, and listening to the interviews), which I conduct as a teacher to gain a provisional understanding of the results of the ProE model implementation. The second is the systematic analysis of this investigation, to which this section refers.
Thus, thematic analysis, as defined by Braun and Clarke (2006: 6), “is a method for identifying, analysing, and reporting patterns (themes) within the data”. Therefore, in this method the researcher essentially immerses him/herself in the qualitative data in order to capture key aspects relevant to the research question(s) (Vaismoradi et al.,
2013). These aspects are then grouped into themes, which are ordered, validated, and analysed in order to produce a report that advances understanding of the studied phenomenon.
Thematic analysis offers a rich array of advantages over similar methods (e.g., content analysis), which is the main reason I opted for this particular method of data analysis. Firstly, thematic analysis is highly recommended for novice researchers like myself because it does not require specific theoretical or technical skills. In fact, as Vaismoradi et al. note (2013: 400), it actually “provides core skills to researchers for conducting many other forms of qualitative analysis”. Therefore, with this analytical approach, I am less likely to provide a misleading or deceptive account of the data due to my relative inexperience as a researcher. A similar advantage is that thematic analysis is versatile in nature, which makes it appropriate for a wide range of epistemologies and research questions (Vaismoradi et al. 2013). As Braun and Clarke state (2006: 5), “through its theoretical freedom, [thematic analysis] provides a flexible and useful research tool, which can potentially provide a rich and detailed, yet complete account of the data”. In view of this, thematic analysis seems an essential method to master for my current and future career as a researcher given its applications in different research situations.
In another vein, I would argue that another advantage is that, unlike content analysis, thematic analysis not only looks into what recurs most in the data (i.e., quantification of data) but also considers what might appear relatively insignificant (Spencer et al., 2003; Vaismoradi et al., 2013). As Braun and Clarke (2006: 10) observe, in thematic analysis, “the 'keyness' of a theme is not necessarily dependent on quantifiable measures — but in terms of whether it captures something important in relation to the overall research question”. On another note, thematic analysis also permits the researcher to link the analysis of meaning to the context in which the data is generated (Joffe & Yardley, 2004). This means that in contrast to, for example, conventional content analysis (see
Hsieh & Shannon, 2005), whereby the number of instances across the data is the major determinant of significant meanings (Morgan 1993; Twycross & Shields, 2008), in thematic analysis, specific contexts largely contribute to shaping significant meanings. All this is particularly relevant to this investigation as, firstly, I wish to explore frequent occurrences (i.e., the expected) in addition to what is significant (i.e., the unexpected). Secondly, this being an AR(-like) study that seeks deep understanding, as opposed to qualitative studies that focus on explanation, context warrants special attention if I want to accurately identify, analyse and interpret my research topic.
Having said this, despite all its advantages, I am also aware of the disadvantages that thematic analysis poses — which, as Braun and Clarke (2006: 27) note, “depend more on poorly conducted analyses or inappropriate research question than the method itself”. For example, it is often reported that thematic analysis is more descriptive than
interpretative (see Braun & Clarke, 2006; Vaismoradi et al., 2013, for a discussion). The reason for this seems to be that it is generally regarded as having a somewhat inductive (data-driven) nature that can hardly move from describing characteristics of a
phenomenon to describing how, when, or why those characteristics occurred.
Nevertheless, as Braun and Clarke (2006) suggest, this disadvantage can be easily offset by anchoring the data analysis to an existing theoretical framework or, in other words, using a predominantly deductive approach. Thus, in order to increase the interpretative power of this method I carried out the coding process from a primarily deductive or theory-driven approach (see Elo & Kyngäs, 2008; Hsieh & Shannon, 2005; Vaismoradi et al., 2013). This form of thematic analysis, “tend[s] to be driven by the researcher's theoretical or analytic interest in the area, and it is thus more analyst-driven” (Braun & Clarke, 2006: 12). Consequently, in this study I code the data for the research question, trying to fit the codes, in most instances, into coding frames derived from existing theory. In addition to this, where necessary, I supplement the deductive approach with a data-driven or inductive approach, meaning that I also embrace emerging patterns even
though they have little in connection with existing theory (Kondracki et al., 2002; Mayring 2002).
My approach to the data analysis, then, is from the perspective of coding for the research question. To do so I rely on existing theory as well as on the meaningful insights that the data supplies, regardless of whether these fit into pre-existing coding frames. I suggest that a theory-driven approach, supported by a data-driven approach, should allow me to overcome the drawbacks of each approach and also increase trustworthiness (Braun & Clarke, 2006; Vaismoradi et al., 2013). To clarify, at times, inductive approaches may lead to loss of focus and direction due to focusing only on the emerging areas (see Shamsini 2012, for a discussion). Here deductive approaches can help connect all the emerging themes. Furthermore, in deductive analysis, according to Hsieh and Shannon (2005: 1283) “[…R]esearchers [can] approach the data with an informed but,
nonetheless, strong bias. Hence, researchers might be more likely to find evidence that is supportive rather than nonsupportive of a theory”. Moreover, “an overemphasis on the theory can blind researchers to contextual aspects of the phenomenon” (ibid.). An inductive approach, then, should be useful in allowing themes to flow from the data, which, in turn, should lead towards a richer and more credible analysis.
In view of the preceding, for the analysis process I follow Braun and Clarke's (2006) six phases of thematic analysis (Table 4.5). It is noteworthy that these phases mostly refer to inductive analysis approaches, since they mainly help researchers extract final themes from the raw data. Naturally, where the themes are already established by, for example, theoretical frameworks, some of the phases of the list can be omitted. That being said, I would like to state that four different analyses are conducted to answer this study’s research question (i.e., the impact of the ProE model on student ASEC engagement) in each of the study’s two research cycles. Three of them are deductive (i.e., global analysis of ASEC engagement, temporal analysis of ASEC engagement, and analysis of
the facilitators of engagement) whereas the remaining one is inductive (i.e., analysis of the inhibitors of engagement).
Table 4.4: Braun and Clarke’s (2006) steps of thematic analysis
1. Transcribing and familiarising with the data
2. Generating initial codes
3. Searching for themes
4. Reviewing themes
5. Defining and naming themes 6. Producing the report
Braun and Clarke's (2006) six-phase approach to thematic analysis is a comprehensive and rigorous method for conducting data analysis. It treats analysis as an on-going organic process in which recurrent immersions into the data from diverse angles are required (i.e., comprehensive). Consequently, through this method it is also possible to provide a systematic, consistent, and coherent account of the data (i.e., rigorous) — whose absence in qualitative methods is much criticised (see Holloway & Todres, 2003 and Vaismoradi et al., 2013, for a discussion). By adopting this multi-level approach to thematic analysis I should be able to produce a high-quality report that would hopefully provide new knowledge and understanding.
Thus, I would begin by transcribing the verbal (interviews) as well as the written data (open-response questionnaires). With the latter I mean digitising and preparing the handwritten texts produced by the students for subsequent analysis. For the transcripts, since this is not an investigation that focuses on language use or on speech functionality, but rather on meaning as already discussed in 4.4.1.4, I would concentrate on providing a thorough, orthographic account of all verbal utterances, which shall represent the
original source as accurately as possible (see Appendix B). Following this, I would then familiarise myself with the data. In Braun and Clarke's (2006) opinion, this means repeatedly reading the previously transcribed data and note taking. This step would allow me to obtain an overview of the data, thus guiding me through the rest of the analysis (Polit & Beck, 2004).
After gaining a general idea of the contents of the data, I would move on to produce the initial codes, or groups of meaningful information, as Boyatzis (1998: 63) defines them. As my approach would be both (predominantly) deductive and inductive, I would code for the research question using a crosscutting approach that would go beyond my theoretical stance. This means that in addition to coding driven by the theory I would also generate codes for salient features of the data even though they harmonise little with the theoretical approach directing the data analysis (Braun & Clarke, 2006; Hsieh & Shannon, 2005; Vaismoradi et al., 2013). Finally, the coding would be done using the software programme NVivo (see Kelle 1997; Seale 2000, for a discussion) as this seems to be the leading software for qualitative data analysis and also the one for which training is available at the University of Warwick (see Illustration 4.1 below).
Illustration 4.1: Example of coding using NVivo
Having coded the data, the next stage would be to search for themes. This step would entail generating a provisional list of potential themes and sub-themes from the identified codes, which would be organised into a hierarchical structure (Braun & Clarke, 2006; Morse & Field, 1995) that would be refined in the following stage. It is noteworthy that this stage will be omitted in cases in which the codes, and therefore the themes are already provided by existing theory. Next, the fourth stage consists of fine- tuning the different themes until a thematic map is devised. This would be carried out at two levels. Firstly, I would read all the extracts under each theme to identify potential, coherent patterns. In the absence of such coherence, I would rework the theme until its extracts fit with the idea portrayed by the theme. Secondly, once I am satisfied with the themes, I would go through the entire data set in order to ascertain whether the themes accurately represent the contents of the data. Similarly, in the event that the themes are not coherent with the data I would further review and refine my coding. Finally, I would produce a thematic map of my data, which, although incomplete, should satisfactorily represent the data (Braun & Clarke, 2006) (Illustration 4.2).
Illustration 4.2: Initial thematic map in NVivo
The next step would be to define and label the themes, especially those that emerge from the data. For this, I would again go through the extracts in order to see whether the working names given to each theme accurately represent the subject of each theme. Here I would make sure that labels are not too broad, too diverse or too complex to
understand (Braun & Clarke, 2006). As part of this process, I would also define and label the sub-themes using the same method described above. Having done that, I would provide the final labels for my themes, which should be clear, sharp and straight to the point (Braun & Clarke, 2006).
The final step of thematic analysis is to produce the final report. According to Braun and Clarke (2006: 23), this consists of “tell[ing] the complicated story of your data in a way which convinces the reader of the merit and validity of your analysis”. Accordingly, in my reporting of the data I would always try to use vivid and clear extracts from my data to illustrate my points and/or themes (Braun & Clarke, 2006). Furthermore, to avoid engaging in the act of cherry picking (see Morse 2010, for a discussion) and to thus increase the credibility of my reporting, when possible, I would use more than one extract from different sources (Shamsini 2012); I would account for the deviant cases encountered in the data (Hsieh & Shannon, 2005); and I would use appropriate
(Braun & Clarke, 2006). Then I would weave all of these together in an analytic narrative, relevant to my research question, which would go beyond a mere description of the data (Braun & Clarke, 2006). Following the aforementioned steps, then, should allow me to provide a convincing narrative illustrating the data’s veracity.