Terms used in models related to end-of-life activities
5.4 Analysis of transcripts 1 Method
Transcribed interviews were qualitatively analysed through a thematic coding method. In this context, coding is described as the “process of focusing a mass amount of free-form
data with the goal of empirically illuminating answers to research questions” (Hahn, 2008,
p. 5). A code in qualitative inquiry is “most often 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, 2013, p. 3). Coding techniques are typically
applied to help organise and analyse the considerable amounts of data (in the interview transcripts) that are frequently collected during qualitative research (Holton, 2010).
In this study, developed codes were used to segregate, classify, and group link data iteratively as patterns or themes were identified, and meaning and explanation emerged from analysis, using a form of template analysis113. This form of qualitative data analysis is
favoured by those who find other methods, such as grounded theory, too restrictive, as it allows a degree of tailoring to meet the particular requirements of a study. Template analysis can also handle larger dataset than IPA114, for example with the latter typically
having less than 10 samples, while 20-30 would be common for the former (King, 2004, p. 257). Of course, there are well in excess of 30 interviews in this study, however the flexibility of template analysis allows it to handle larger databases, although this can be to the detriment of developing a particular understanding of an individual case (Brooks, McCluskey, Turley, & King, 2015). As the focus of this study is as much across case as within a case analysis, that is not a significant issue.
There are six steps in the procedure for carrying out template analysis, viz.
113 As discussed on page 44, Template analysis comprises a group of techniques for organising and
analysing data, rather than a specific method (King, 2004, p. 256).
• Familiarise oneself with the accounts to be analysed. This may involve reading all accounts where the number is relatively small or a selection if the number is larger (Brooks et al., 2015); • Carry out preliminary coding of the data. This is similar to other qualitative analysis techniques; however, template analysis permits a priori coding (i.e., from deduction rather than from observation). The interview schedule can provide a starting point for such codes (King, 2004, p. 259); • Organise emergent themes into meaningful clusters, and explore how they relate to each other, both within and between these groupings. This can include both hierarchical relationships with broader ones, as well as lateral relationships across clusters (Brooks et al., 2015); • Define an initial coding template. This is typically done using a small sub-set of the sample. King (2004, p. 259) reports using just two accounts for a this task, and notes the advantage of involving others at this stage – if for no other reason than to “force the researcher to justify the inclusion of each code”; • Apply the initial template to further data and modify as the initial codes appear inadequate. Modifications may include insertion, deletion, changing scope, changing higher order classification (King, 2004, pp. 261–262); • Finalise the template and apply it to the full data set. Brooks et al. (2015) observe “In some respects it should be said that there is never a ‘final’ version of the template, in that continued engagement with the data can always suggest further refinements to coding”, while King (2004, p. 263) comments that “one of the most difficult decisions” is to know when to stop developing the analytical template.
5.4.2 Implementation
Organisation
Due to the large number of accounts, a similar division of work was implemented. The author and the UCC team analysed the transcripts arising from the interviews in Ireland, Denmark, Germany and the UK, while the other transcripts was analysed by the local partners under the direction of the author. Prior to starting the analysis, a one-day coding workshop, led by UCC, was held to coordinate activities and to instruct the local partners in coding and in the use of the NVivo software115. To ensure consistency of approach, there
was constant communication during the analysis, including a question and answer mailing forum which was maintained throughout the coding and analysis process, to highlight challenges encountered, potential pitfalls and solutions to common problems.
Coding
In keeping with the template analysis procedure detailed previously, the first task was to become familiar with the accounts to be analysed. To this end a selection of transcripts representing different stakeholders and different countries was read. This (re)familiarisation with the material was a necessary precursor to planning the coding, and a useful starting point for the preliminary coding. As expected, the principal headings in the interview schedules provided the initial a priori codes viz. building/project, energy use, energy renovation, marketplace, stakeholders and sustainable development.
Séror (2005, p. 323) observes that “Good qualitative research involves meticulous data
sorting and organization and carefully using ideas generated by the data”. NVivo software
from QSR was used in this research (see Bazley, 2007). The software is designed to help organise, analyse, and find insights in qualitative data such as interview transcripts. Within NVivo, coding takes place through use of ‘nodes’, which are effectively containers of information, each containing extracted references and excerpts from the source material,
115 NVivo is qualitative data analysis software, which takes the place of the traditional methods of
highlighting, underlining, and (literally) cutting, pasting, copying, and categorising volumes of printed text.
and labelled to capture the meaning of all constituent reference material extracts. Nodes can also be used to record metadata associated with the interviews e.g., interviewees, organisations, and projects. In this way nodes can be used in two ways, i.e., for both analytical and descriptive functions. However, it should be noted that NVivo is still just a tool for facilitating the mechanical steps associated with the analysis116 – however, the
analysis of the material is still very much in the hands of the researcher. The initial codes were subsequently refined and further developed during the coding of the transcripts. An initial coding template was prepared and shared with all partners to ensure a common starting point. Changes to the codes were periodically communicated and partners exchanged a number of transcripts during the analysis process to ensure the coding was aligned. As the coding process advanced, relationships between the codes became apparent, the template was thus structured in a hierarchical manner denoting such relationships. When coding was completed, a summary of each coding node was written up, highlighting key points from the interviews, supported with quotes. The material was then organised under headings reflecting the objectives of the research.
The actual process of using NVivo can be summarised as thus: (i) import of transcripts into NVivo as ‘sources’, (ii) initial review of source, including using tools such as word search, word cloud, etc., (iii) coding of the transcript, noting emerging concepts, (iv) perform ‘queries’ to ascertain relationships between codes, (v) review and reflect on material, (vi) visualise codes in written and graphic form, and (vii) repeat process in a reflexive manner, refining, rearranging and consolidating codes, developing insights, and exploring the relationships between codes. Figure 38 below presents the final coding template, wherein the various codes developed in the analysis were consolidated and standardised.
116 Séror (2005) gives the example of regrouping an informant's data in one folder or category