Chapter 3: Methodology 98
5. Analysis process and theory development 152
5.2 Analysing the data 154
In this section, I will explain the steps taken to analyse the data, beginning with thematic analysis, deeper exploration using line-‐by-‐line coding, improving trustworthiness by approaching the data a third time using a mindmapping technique and additional graphical software, and finally, using a theoretical framework to create a composite analysis.
There are obviously a large number of ways of analysing qualitative data,
and no single ‘correct’ way. Rather, the goal of a qualitative analysis strategy should be to impose systematic order on the data, immerse oneself in it, and align the analysis with the overall research questions and epistemology underlying the project. This comprised a number of steps that were a combination of pre-‐defined process and evolution as the needs of the project were made increasingly apparent.
Step 1: Generating initial codes in thematic analysis
Thematic analysis (TA) is a:
‘poorly demarcated, rarely acknowledged, yet widely used qualitative
analytic method…’ (Braun and Clarke, 2006, p.77)
In it, the researcher seeks the themes in the data. As such, this method is not bound to philosophical perspective (in contrast to, say, Interpretive
Phenomenological Analysis, which is aligned with phenomenology). This lack of
demarcation means that the conscientious researcher must be scrupulously detailed in outlining the method used to generate the themes.
This project used ‘theory-‐led’ TA (Coolican, 2009) as it drew on existing theories surrounding social cognition and adolescence. It did, however, also include an element of ‘inductive’ TA, as the technological element was explored without a preconceived conceptual framework. As a result of this tension, I began the coding process by completing an initial axial coding of each tool in NVIVO. I
used pre-‐determined themes as codes as well as themes emerging from the data. The pre-‐determined codes were high level, related directly to the research questions and without granularity:
Attachment Mentalizing/ToM
Attribution Self
Identity Risk
Technology Adolescence
Axial coding has been criticized for distracting researchers from the themes emerging from the analysis (e.g. Glaser, 1992). Given the conceptual framework within which this project was operating, I did not see that an open coding approach would be a productive first step.
Step 2: Deeper exploration
Once the data set had been reviewed and broadly coded, I revisited each data source in NVIVO, and carried out ‘line-‐by-‐line’ coding, during which a significant number of codes and sub-‐codes emerged alongside the initial axial codes. A full list of codes and sub-‐codes may be found in Appendix 5.
In parallel, during this step I explored the data emerging from the surveys. These findings had to be handled carefully to ensure they were used to illuminate
the research questions in tandem with more traditionally qualitative data, rather than to over-‐claim or claim generalizability for the population.
Step 3: Increasing trustworthiness
It became clear at this point that despite the repeated revisiting of the data, and the attempts to be reflexive in that process (described in more detail in Section 6 of this chapter below), I needed to introduce an additional layer of analysis to provide another ‘way into’ the data, away from NVIVO. This would have the important added benefit of providing an additional layer of robustness; if the themes and findings emerging from this new coding correlated with those from the work in NVIVO, I could be increasingly confident that the findings were reflective of the reality. This explanatory framework resulting from these
additional reflections should, insofar as possible, be ‘saturated’, as a grounded theorist might put it – that is, additional data should not alter the themes and findings emerging. In order to do this, I used two techniques.
Firstly, I once again revisited each data source, and focused this time on
emerging findings that responded to each research question rather than the axial codes I had used in the previous step. I used the Mindnode mindmapping software package to represent my thinking, an example of which may be found in Appendix 6. To use Mindnode I chose a main node from the NVIVO analysis and placed it as the core node on the mindmap map. As important quotes from the data emerged I added them to relevant edges (linking nodes) of the map. Sub-‐nodes in NVIVO correlated well with radial edges in Mindnode.
Secondly, when the analysis called for it, I used yEd to create diagrammatic representations of the data using nodes to represent key points, with edges linking the nodes. yEd is graphical software, intended to enable the user to create and recreate a range of representations of a single data set. In yEd the graphical user interface (GUI) allows the user to choose nodal points and designs, and then to
link them using edges by clicking from one node to another. Edges can be bi-‐ directional, uni-‐directional, or not indicate direction. As before, nodes correlated with main nodes in NVIVO. I used yEd because it offers a range of layout
algorithms enabling creative representation of the data, and highlighting relational aspects of the data that I might otherwise have missed. Once the data have been inputted, the user can easily apply a layout algorithm from the list of hierarchical, organic, orthogonal, circular, tree, radial and series parallel. For each of the concept areas I implemented in yEd I tried all of these algorithms before choosing the representations that were most illuminating and honest.
Step 4: Composite analysis, or bringing it all together
The next stage was to bring the analyses together in a way that responded to the research questions. The goal was to extend the thematic analysis by placing the data within a theoretical framework. The framework needed to (Damasio, 2012):
-‐ define and organize the resolved themes in such a way that produced a strong, complete theoretical response to the research questions
-‐ propose explanations for the findings at the level(s) at which these explanations could apply and make explicit the interconnectedness between levels
-‐ identify the findings amenable to analytical methods available to social research, and resolve discrepancies arising from the previous analytical
stages
-‐ result in messages for young women and the adults in their lives (see Chapters 5 and 6).