Visualisation during coding

Using visualisation to enhance the coding process I recently attended the AASR workshop conducted by Professor Barbara Bowers (University of Wisconsin) on the principles and practice of grounded theory and the AQR presentation by Dr. Helen Marshal (RMIT University) on the challenges faced by those who are coding utilising qualitative analysis software. My purpose was to see how others were using visualisation methods to enhance the coding process in their research. It was clear from those I met, that only a few had actually employed visualisation methods. A large number of researchers indicated they were using NUDIST software that does not have the facility to create visualisations from the textual analysis (coding) undertaken. I have found the use of the visualisation features of ATLAS.ti software valuable in coding process. Before I give you my account of visualisation process I used during coding, I will give you an introduction into my research to put my understanding of the visualisation process into context. My PhD research is an examination of current engineering management practices in infrastructure organisations, particularly where infrastructure managers claim they apply facilities management, asset management and/or systems engineering methodologies. The underlying basis for my research is the application of grounded theory. The basic building blocks (Pandit, 1996) for the coding process I used were: Concepts: Categories: Relations:

Basic units of analysis; it is not the data itself but the conceptualisation of the data that the theory is developed; Represents grouping of concepts; represents a higher order and more abstract; and provides the means by which the theory can be integrated; Indicates generalised relationships between categories and its concepts, and between categories themselves.

For the purpose of my research, data was analysed “manually” and the software tool used to facilitate/aid this process was ATLAS.ti by Scientific Software. I used the software to:  Visualise complex properties and relations between objects;  Integrate the concepts;  Explore the relationships; and  Go beyond the boundaries of current infrastructure management thinking (Thomas, 1997) Visualisation in this case is done through the use of network views. Networks views are graphical representations that in essence, are based on systems theory. The views are made up of nodes and links. The nodes represented categories, codes, memos, etc. Links were graphically shown as lines or arcs that connect nodes. ATLAS.ti allows these links to be named to indicated the nature of the relationships between nodes (these are called relations). ATLAS.ti has some in-built relations (for example, is associated with, is part of, is cause of) and allows the development of new relations that reflect the nature of the emerging theory. Network views were made continuously during the life of the research so as to make the many connections among major elements of the emerging theory visually and conceptually clearer .

The set of relations used in this research was a balance between those used in developing models in the systems engineering software (such as CORE) and those most commonly used in the social science context (see to Table 1). The first step of my research was to interpret of data gathered from engineers, managers, stakeholders, professionals etc. who influence the technical management practices of infrastructure systems. These interpretations were used to generate concepts, which refer in this context, to an abstract representation of an incident, idea, event, object, or action/interaction that is judged significant in the data. These provided a basis to construct an explanation that validated the interpretations of the observations. Relation relates part cause contradicts determine evidence impacts issue

Explanation Relates concepts without assigning a rule. Identifies the domain set by this element. The part-of relation links objects, not concepts of different abstractional level. Used for representing causal links, processes, etc between elements. Indicates inconsistency or opposition between elements, identifying differences. Where one element regulates changes in the other (control). Identifies information that serves to support the concept. Identifies a strong effect. Identifies a problem with any element in terms of completeness, consistency, correctness, testability, etc.

Table 1: Extract from List of Code Relations Relations such as these are used to describe the relationships between codes, and are stored as global relations in ATLAS.ti. When new network views are generated, relations tagged to codes reappear if the same codes are used again. This is an important feature as it allows the examination of the relation in a different context and a means to compare with previous network views (the relation gets ‘carried over’ to the next view). This feature of ATLAS.ti has its downside, however, as it is unable to filter out specific relations in the network views. That is, all relations are shown between codes in the network views. Codes and their corresponding relations can be taken from the view, but relations themselves cannot be isolated from view. As new network views are constructed, they generally represent different perspectives or contextual meanings and as such generate more relations. There is no option to hide particular relations that are not relevant to the perspective under review. As I developed network views, the relations drawn from other perspectives are included in the network view, making them unnecessarily complex. Deleting any relations from the network view results in the global elimination of the relation from the analysis.

An example of a network view generated using ATLAS.ti is shown in figure 1 below:

There are a number of ways to address this issue, for example, opening up a new file for every document to interpret the relations (hermeneutic unit) and merging the files into one file at a later stage. This did cross my mind but I felt that working through the documents in a single file provided more advantages than disadvantages, as the nature of the grounded theory process is to focus on grouping codes (by the formation of categories) and examining these at a higher conceptual level. At this level, relations can be generalised to represent the underlying variations of behaviour identified during the early stages of the analysis. The use of grounded theory as described by Strauss & Corbin (1990) enabled me to capture the complexity of reality (phenomena) under study and to make convincing sense of it. I placed particular importance on the role of the development of network views – the use of mind visualisations or system maps – to gain greater insight to the relationships that mean most to those involved in the management of infrastructure. Visualisation enabled the research to exploit the theoretical models and the underlying structural properties in a way that far extends the boundaries of textual analysis. Establishing and refining the relationships between categories and underlying concepts, and between categories themselves, was the key benefit of developing network or semantic views (for example, the visual representation of codes and relations). It was evident during the coding process that it was necessary to have flexibility in representing the data and sets of data (seeking different ways to view the data). This allowed opportunities to deal with the confusion that often presented itself (reducing uncertainty) during the coding process and for new concepts to develop.

References: Muhr, Thomas. ATLAS.ti Version 4.1 for Windows 95: Visual Qualitative Data Analysis Management Model Building in Education Research and Business: Short User's Manual. Ist. Scientific Software Developments, November 1997. Accessed September 1998. Acrobat Format. Available from http://www.atlasti.de/. Pandit, Naresh R. The Creation of Theory: A Recent Application of the Grounded Theory Method. Volume 2, Number 4. The Qualitative Report, December 1996. Accessed June 30 1998. HTML. Available from http://www.nova.edu/ssss/QR/QR2-4/pandit.html. Strauss, Anselm, and Juliet Corbin. Basics of Qualitative Research: Grounded Theory Procedures and Techniques. Newbury Park, California: SAGE, 1990. Author Background: Ralph I. Godau Ralph graduated from RMIT University in Civil Engineering (1976) and continued his education by completing a postgraduate studies in Business Administration (1984) and a Masters in Systems Engineering (1995). Ralph has also completed a Graduate Certificate of Tertiary Teaching & Learning (1997). Ralph worked for a gas distribution company for 18 years and was involved in many facets of the organisation including asset management, information technology, systems engineering, project management and day labour management. Ralph is currently working full-time for the Office of Gas Safety, a regulatory body responsible for assuring that safety management systems are in place to adequately meet the needs of the community. Ralph is near completion of his PhD at RMIT University in the area of Infrastructure Management and is employed part-time as a Systems Engineering and Engineering Management sessional lecturer. Ralph has activity been involved in facilitating learning at RMIT University for over 6 years.