• No results found

3 Research Design and Approach

3.7 Data coding and analysis - NVIVO

The qualitative analytical tool, NVIVO, was employed to code the data based upon the categories noted previously. The author could then more easily look for significant statements and recurring themes therein. This information could then be subdivided to give a textural (what is being experienced with respect to LO processes) and structural (how are the processes being experienced situationally and contextually) understanding of the individuals’ experiences.

At the outset of the research, it was not assumed that the data collected would triangulate and support the same conclusions. Whilst data convergence was anticipated, there was the possibility that the evidence uncovered from the various

data sources would be divergent. This was an issue that the author considered at an early stage and was ready to deal with as the research progressed.

The underlying strategy for the coding and analysis of the data was that of using the current theory to provide the framework (Flick, 2006; Yin, 2003) in a deductive approach. As can be seen from the areas of questioning, there already exists a theoretical model about the potential processes which influence performance taken largely from Senge et al (1990, 1994) and subsequent research stemming from this seminal work. Yin (2003) states that perhaps the most difficult elements of the case study approach are the coding and analysis of the data collected. As such, he proposed a simple initial approach taken from Miles & Huberman (1994) which contains six steps to get the data into order for further analysis to take place. It involves putting the data into different arrays and then creating a matrix of categories within which to place the various data collected.

The theoretical headings which were used to formulate the interviews form the basis of the data analysis matrix and flowcharts. The data collected through the case study research was coded into the headings discussed in the literature review section. This approach allowed a code to be developed before commencement of the field research itself. In addition, such an approach helped keep a consistent framework running through the work. A deductive approach was chosen over an inductive approach for several reasons: firstly, given that the hypothesis was to test the presence of what was already a relatively well researched phenomenon in industries outside construction, it was felt that in order that this thesis could provide a level of comparison for the operation of the LO model in construction against other industries.

Secondly, this thesis was based upon a theoretical discussion between the researcher and one of the participant customer organizations and therefore the research herein reported started with a theory and a hypothesis as its starting point.

It is accepted that part of the field research behind this thesis involves the observation of the operation of processes within the subject organizations, and therefore there is an argument that an inductive research methodology would have been an equally valid approach. An inductive approach does have the advantage of being open ended and allowing greater exploratory freedom, but was rejected based upon the lack of key advantages, set out in the paragraph above, which a deductive approach provides.

The risk with taking a deductive ‘pre-coding’ approach, however, was that the author may try to shoehorn data into one of the LO process headings rather than accept early on that there may be other factors at work (Yin, 2003). This risk was reduced by the fact that the theoretical framework was partly based upon existing research which already suggested the existence of the processes under study (Bhatnagar, 2006; Garcia-Morales et al, 2007; Skerlavaj et al, 2006). Following the establishment of a theoretical framework, the key categories were then placed into flowcharts and interdependency matrices and the frequency of occurrences of these themes could be mapped onto these charts.

NVIVO was used as the primary case study database as much of the data collected was qualitative in nature. NVIVO lent itself to easy coding and recoding as necessary as the data were collected. The bulk of the qualitative data were the interview transcripts and documentation, which could be coded relatively easily. Therefore the content of NVIVO was coded not only by the subject headings noted earlier, but by the source of the information itself. This aided efforts to triangulate the data collected within the theoretical framework.

Yin (2003) states that the database is an often ignored element of case study research which provides the background evidential data for the final report. In addition, the principle of being able to provide a chain of evidence which can be followed by anyone reading this work is adhered to as NVIVO allowed all data for this thesis to be visibly coded and linked (Flick, 2006; Silverman, 2006). Using NVIVO to provide this chain gives this thesis more visible credibility when read. Credibility is another key criticism of case study research (Yin, 2003) and the good documentation of the processes and research material should allow someone else to replicate this research should they so choose.

The author must also accept that research bias is inevitable within this work. The author takes the position that it is impossible to be completely objective about all aspects of the research. In addition, the mere fact that a researcher interacts with individuals during interviews and focus groups may mean that they are influenced in some way. This influence may be seen to be particularly evident as the interviews are not conducted in a fully structured paradigm (Flick, 2006; Silverman, 2006). The aim of this aspect of the research is always to remain as independent from the subjects under study as possible.

3.8 Research design and approach leading into data collection and