4.5 Choosing a data collection framework
4.5.2 Concept of content analysis
Qualitative research ‘very rapidly generates a large cumbersome database’, due to extensive interview transcripts and field notes, which can result in the researcher feeling lost or swamped by it all, but ‘it is crucial to guard against failing to carry out a true analysis’. There are few well-established and widely accepted rules for the analysis of qualitative data’ (Bryman, 2008, 538). There is no codification structure as
175 such for qualitative data analysis, but ‘many writers would argue that this is not necessarily desirable anyway’ (Bryman, 2008, 538). There are broad guidelines to qualitative data analysis. All the research data has been analyzed and evaluated through a content analysis method.
Coding is a form of ‘analytic scaffolding’ which research is built on (Charmaz, 2005, 517). Researchers study their findings meticulously to find new leads and ‘gaps in them. Each piece of data ... can inform earlier data’. Therefore researchers who develop ‘a code in one interview’ can re-check previous interviews for similar codes (Charmaz, 2005, 517). This approach is based on the concepts of grounded theory although it may be used in coding content analysis in comparing ‘data with data, data with categories, and category with category’ (Charmaz, 2005, 517). The purpose of coding is to be able to develop similar themes and patterns that will form the foundation for the interpretations in the research. As a researcher working on a specific piece of research, it is also important to ensure that the coding system critically analyses the research data and that the coding scheme is the researcher’s own. This ensures that the research is more competent (Rapley, 2011, 283). Content analysis allows for the categorisation of the data into themes and similarities and subsequently a system of coding.
Content analysis is a very transparent research method. The coding scheme and the sampling procedures can be clearly set out so that replications and follow–up studies are feasible. It is this transparency that often causes content analysis to be referred to as an objective method of analysis
(Bryman, 2004, 195).
Coding the meanings into categories allows the researcher to ‘quantify how often specific themes were addressed in a text, and the frequency of themes can then be compared and correlated with other measures’ (Kvale, 2007, 105). It is important that the method of analysis permits the researcher to ‘maximise the potential for a full and reflective analysis...and allow emergent ideas, concepts, patterns etc. to remain rooted within original data’ (Spencer, Ritchie & O’Connor, 2003, 217). This is an important aspect of the research process. The researcher needs to be aware of the validity of the categories and themes. There are four factors that can influence validation, frequency, specificity, emotion and extensiveness, (Krueger & Casey, 2000, 136). This part of the
176 process allows the researcher to be more critically aware of what is or is not important about the data. Researchers take note of how frequently something is said but ‘it is a huge mistake to assume that what is said more frequently is most important. Sometimes a really key insight may have been said only once…You have to know enough about what you are studying to spot a gem when it comes along’ (Krueger & Casey, 2000, 136). The concept of ‘specificity’ suggests that researchers tend to lay more emphasis on feedback that is specific, (2000:136). The concept of ‘emotion’ suggests that researchers tend to pay more attention ‘to comments or themes in which participants show emotion, enthusiasm, passion, or intensity in their answers’ (Krueger & Casey, 2000, 136). The final concept is extensiveness, which has similar traits to frequency but is slightly different. Extensiveness refers to how many different people said something. Frequency is how many times something is said’, but this could mean one person repeatedly bringing it up; therefore awareness of extensiveness may be warranted in the research, (Krueger & Casey, 2000, 136).
Ritchie, Spencer and O’Connor, (2003) describe the linkages found between sets of data as ‘matched set linkages’ (2003, 248-249). In other words, the attitudes of one sub- section of the data, may agree, in that they all disagree, in response to something, but the reasons may differ as to why they disagree. In data management, it is only by scrutinising the data intensively that the ‘lines of enquiry to pursue, or the puzzles posed by the data, begin to emerge’ (Ritchie, Spencer & O’Connor, 2003, 261). One common criticism of coding relates to ‘the possible problem of losing the context of what is said’ (Bryman, 2004, 411). A researcher needs to be very particular about this part of the process in content analysis, as it is important to retain the voice of the research participants, since it is their voice that lays the foundation of the research. Many researchers often feel inundated at this point in the research, as there may be so many codes and interrelated categories that they feel overwhelmed. Bryman (2008) approaches the process step by step. He suggests transcribing interviews as they are conducted, as this is the start of a framework for coding, and ‘may sharpen your understanding of the data and help with theoretical sampling’ (Bryman, 2008, 55). Theoretical sampling is a concept of grounded theory and it refers to ‘sampling carried out so that emerging theoretical considerations guide the selection of cases and/or research participants’ (Bryman, 2008, 700). At this stage in the analysis, the researcher should be thinking about developing general theoretical ideas in the data, (Bryman,
177 2008, 550-552) and developing linkages. Essentially, the researcher is refining notes into codes, an approach also known as the ‘long table approach’ where notes were cut and pasted into groups and themes (Krueger & Casey, 2000, 132). However, Bryman suggests that computer software is now increasingly being used to perform these tasks (2008, 552).
Initially, the researcher began coding, using NVivo 8 as a replacement for the ‘long- table approach’ as a cutting and pasting tool but had to revert back to the ‘long-table approach’ when working from a different computer which did not have the NVivo8 programme on it. NVivo is one of the most commonly used computer software packages (Barbour, 2008, 195). Computer ‘software warrants serious consideration because of its power and flexibility’ (Bryman, 2008:582), even though criticisms have been raised that qualitative software packages will become more like quantitative research in that it creates a temptation to quantify findings (Bryman, 2008, 566). There is also a fear that the narrative flow of interview transcripts and events recorded in field notes may be lost, as the software reinforces the ‘code and retrieve process’ (Bryman, 2008, 566). It has also been argued that the flow of communication between focus group participants could be mislaid in the code and retrieve process. On the positive side, it has been argued that CAQDAS may be helpful in the development of explanations, such as age gender and so on. Software programmes like NVivo create the opportunity for the researcher to ‘think about codes that are developed in terms of ‘trees’ of interrelated ideas...and [this] urges the analyst to consider possible connections between codes, (Bryman, 2008, 567). Many researchers use both methods: computer software and pen and paper (Barbour, 2008). The interview process also involved interviews conducted over the telephone, which will be discussed in the following section.