NVivo is a computer package that can facilitate the analysis of qualitative data (Bazeley & Jackson, 2013). It enables the researcher to deal with voluminous, text- heavy and cognitively demanding data sets as it allows for data to be stored, updated and changed in electronic folders (Liamputtong, 2009; Yuen & Richards, 1994). NVivo does not conduct data analysis for the researcher, unlike statistical computer packages such as SPSS, but offers an organisational package that can assist the researcher with coding and categorising the data (Liamputtong, 2009). For example, the interviews for this thesis produced over 494,000 transcribed words. Although not impossible, hand coding such a large data set would have been very demanding on the researcher who would have to manually search between interview transcripts when referencing common themes. NVivo facilitates electronic search of transcripts, which is much faster and lessens demands on the researcher. Although NVivo is primarily used to analyse interview transcripts, it can also facilitate the analysis of other types of qualitative data including picture files and audio files (Bazeley & Jackson, 2013; Reavey 2011). The remaining section of this chapter will detail how
53 NVivo was utilised to facilitate the GT analysis of CDM interview transcripts in this thesis.
2.6.1 Creating a project in NVivo: Preparing data for analysis
The first step in qualitative data analysis involves preparing the data for analysis (e.g. transcribing). When using NVivo, this means creating a ‘data project’. A data project is a computerised file in which all information about the project will be stored; much like a folder on a desktop computer. Not only does it store raw data and subsequent sorting and analyses, but it can also include documents that can help audit the research process. For example, it is recommended that researchers who use NVivo create an electronic ‘journal’, which is stored within the data project (Bazeley & Jackson, 2013). This acts an electronic notebook where the researcher can document their thoughts, reflections and actions during the analysis process. It serves as a useful audit trail that the researcher can return to when self-monitoring and grounding their theoretical understanding of the data (Liamputtong, 2009; Mueser and Nagel, 2009). It also provides an appropriate place to store ‘memos’ that enhance the ‘theoretical sensitivity’ of GT analyses (Amsteus, 2014). During data preparation in NVivo, the researcher can also transcribe data from audio to written format. This can be done within NVivo itself or by importing a typed transcript from Microsoft Word. A total of n=31 interviews were transcribed in this thesis, which totalled to 51 hours, 18 minutes and 34 seconds of audio files with a mean length interview of 1 hour, 39 minutes and 18 seconds. During initial transcription for data preparation, the researcher made notes on preliminary emergent themes in preparation for GT analysis and coding.
2.6.2 Coding in NVivo: Creating nodes
Once data has been effectively prepared and organised (i.e. transcribed), it is then possible to begin coding the data. NVivo does not code the data for the researcher, but instead helps to organise codes electronically (Liamputtong, 2009). Codes are referred to as ‘nodes’ in NVivo, which the researcher can create, edit, delete or merge at any point during the analysis process. This was hugely advantageous during the GT analyses performed for this thesis, as it facilitated flexibility when identifying emergent themes. For the current thesis, each transcript was coded one at a time within NVivo; when an utterance within a transcript seemed
54 relevant and interesting (i.e. related to decision inertia), the relevant text was highlighted in the document and a ‘new node’ was created. Whenever a ‘new node’ was created, it was named with an identifiable label (e.g. ‘lack of information’) and a ‘description’ of the node’s meaning was added to the file. The description box helped facilitate ‘constant comparison’ as a part of the GT process, as it was then possible to see how different nodes related to one another later in the analysis process. Once a ‘node’ has been created, subsequent text could be quickly coded into the node by highlighting the text and then dragging and dropping it into the node folder. It was also possible to review node content at any point during analysis by opening the node folder, which contained all the text that had been placed under that code. As nodes were stored electronically, it was possible to create a huge volume of nodes that allowed for wide and diverse coding. It also enabled multiple coding of the same portion of text into multiple nodes when required.
2.6.3 Refining codes in NVivo: Parent nodes
Once the initial coding of transcripts was completed, the next phase of GT involves the refinement of nodes through constant comparison (Amsteus, 2014). In NVivo this process involved the creation of ‘parent nodes’. Parent nodes are top level themes under which relevant codes are stored. Parent nodes effectively act as ‘themes’ in GT analyses, which are broken down into further ‘sub-themes’ and ‘codes’. The creation of ‘parent nodes’ (i.e. themes) facilitates the early consideration of the final theoretical argument; a fundamental aim of GT. NVivo is especially useful for inductive analysis, as the researcher may create ‘miscellaneous' parent nodes under which to store emerging and unusual themes for later analysis. For example, the researcher of this thesis was able to code data relating to both goal orientation (Chapter 4) and uncertainty (Chapter 5) at the same time by creating separate parent nodes that focussed on each of these specific areas.
2.6.4 Knowledge Representation: Concept maps
In addition to facilitating the analysis of qualitative data, NVivo can also assist with knowledge representation via the creation of visual models. Depending upon research requirements, NVivo can produce basic frequency data with reference to both textual data and coding. For instance, it is possible to generate ‘word trees’ that identify the words in the transcripts are commonly linked to one another. NVivo can
55 also visually display the most common nodes within the data set. It is also possible to tag interviews (e.g. Fire and Rescue interviews; Police interviews; Ambulance interviews) and produce frequency data to compare trends across data sets. There is also a wide range of graphic and visualisation tools that can be utilised to build concept maps; a recommended format for CTA knowledge representation (Crandall, et al., 2006). Thus, NVivo is an example of a useful technology that can facilitate qualitative research by enabling researchers to explore large data sets in a robust and systematic fashion.