A. Analisis Univariado
V. Discusión
Content Analysis (CA) is a flexible method for interpreting qualitative data (241, 242). This systematic process condenses large amounts of raw text data and translates it into
categories then themes (241, 242). Reducing the data to categories then themes allows the researcher to establish a clearer picture about what the data is representing. The CA process allows researchers to obtain a greater understanding of social realties in a subjective but systematic and scientific way (242). Like all research paradigms, CA has evolved over time. There are now three variations of CA used to analyse qualitative data: Summative, Directed and Conventional CA (241, 242). Given the varying definitions of CA, it is necessary to describe the process in each research case.
Summative CA analysis begins with counting common words or content (242). Further exploration of the words occurs to understand their usage in context of the research (241, 242). Summative CA aims to understand the contextual use of words. Directed CA begins with an existing theory or research and initial coding of the categories is based on key concepts or variables from the theory in question (241, 242). The researcher then analyses the data using predetermined codes and new categories may emerge (242). Directed CA aims to validate, refute or extend a theory (241).
Conventional CA is used in the absence of an existing theory about a phenomenon and generates new knowledge to develop concepts, ideas, behaviour models or theories (241, 242). The data directly informs the coding process, categories then emerge from the data through inductive reasoning during the analysis (242). The researcher then creates themes from the categories to understand a phenomenon (241); in the case of this research, children’s eating behaviours (241). As there are limitations in the literature about the
phenomenon surrounding children’s eating behaviour, this research did not set out to test an existing theory but rather to generate new knowledge. Therefore, conventional CA was used to analyse data from Action Cycle 3 and 5.
The benefits of Conventional CA is that data collection is from direct participation and without pre-empted theories or categories, which can reduce bias (241). This of course can also be a disadvantage, there is a risk that categories are missed or the link between data and the findings is compromised (241). To strengthen data analysis the researcher incorporated processes known to establish trustworthiness in qualitative research; credibility,
transferability, dependability and confirmability (241-243) .
The basis of credibility is twofold; use an appropriate transparent process to collect and analyse data that will provide an accurate representation of the study participant’s
perspectives; and clearly describe the processes used to analyse the data (242). Techniques used to improve credibility included collecting data over multiple time points, accessing data directly from children in their own environment, using a peer debriefing process during the research (through supervisors, conferences and journal publications), using concepts that emerged from multiple respondents and refining the research as more information became available following each action cycle (241, 242).
Another technique to strengthen credibility was gaining knowledge from different sources at different points in time across five different action cycles, comparing the data to ensure the findings concur across the data and synthesising the information at the end of analysis (243).
Transferability is the extent that the research can be transferred to another context (243). It is the researcher’s role to provide quality data and descriptions, which allow other researchers to make judgement about the application of the research to various contexts (243). The researcher presented the data by documenting the research in a thesis, presenting to local stakeholders, at national conferences and publishing in two international journals. The feedback thus far about the model that was developed from the research is that, the results are clear and potential application of the model into the school setting is easy to understand. Dependability and confirmability reflect the links between research processes, data collection and interpretation of findings (242, 243). Dependability was strengthened by using a
transparent process that was reviewed regularly by supervisors and using inter-coder verification (supervisors KS and LM) to analyse the data (242). Confirmability was strengthened by sharing the research with a range of academics and practitioners in a variety of ways (department meetings, workshops, conferences, journal article submissions), then applying any feedback to research documentation to improve the articulation of the research.
In line with conventional CA as described by Hseih and Shannon (241), and support by Zhang and Wildemuth (242), data was analysed to develop codes and categories (Table 4.2). There were two slight deviations from Hsieh and Shannon’s approach to Conventional CA (241), clusters are referred to as themes in this research and themes were then used to conceptualise the data into theoretical concepts. Over several iterations the coding schema, general categories and then themes emerged. Unlike other versions of CA, the coding schema was not established before the data analysis began. The detailed process used (presented in Table 4.2) was based on Hsieh and Shannon’s work (241):
Video clips were watched and notes taken to allow the researcher to immerse herself into the data and get a sense of the whole.
During the second viewing of the video clips, key words that described reasons why children chose food and other repeated codes were highlighted. During this iteration, the coding units and categories began to emerge.
Steps 1 and 2 were completed for Action Cycle 3 before Action Cycle 5 commenced. From this point in the process, a more detailed analysis was performed on each action cycle individually but with a much shorter period between analysing each cycle.
Video clips that were more than 10 seconds long and relevant to the research were transcribed verbatim for each action cycle.
The researcher fully engaged with the transcripts by reading them word for word, highlighting quotes that captured the thoughts of children and repeated words and phrases, which let codes emerge.
After the transcripts were read again, with further annotation and checking of codes, supervisors viewed and verified the codes.
Codes were developed with each iteration, which allowed categories to emerge, which was also discussed with supervisors.
This same process was completed independently for each Discovery Day. On the second Discovery Day there were worksheets collected. These were used to support the analysis of video clips and transcription data that came from Discovery Day two.
Once sufficient consistency existed for each independent Discovery Day, the themes from the two different data collection days were synthesised together to create theoretical concepts and then a model to convey children’s decision-making criteria for food choice (Figure 4.9 Conventional Content Analysis plan for Action Cycle 3 and 5).