CHAPTER II. METHODS AND DESIGN 61
III. DATA ANALYSIS PROCEDURES 90
The audios from the 30 interviews with the caregivers were transcribed as verbatim as possible. The researcher transcribed three interviews and three home observations and, due to the extensive time required to transcribe each interview and observation, hired three undergraduates as research assistants (RAs) to assist and transcribe the rest.
The length of the recordings and the background noises in the homes posed a challenge when transcribing the data. Typically, all the homes had the TV on all the time and very frequently music was also playing on the radio. In many cases, especially when families lived in an apartment or a public housing block, noises from the neighbouring families were also audible. Two of the RAs resigned after transcribing a couple of cases, so the remaining RA performed 80% of the transcriptions. The researcher checked 25% of the transcriptions back to the audios in order to verify their accuracy.
III.2 Data analysis
The thematic analysis of the transcribed data was performed with the N-‐Vivo software. This researcher developed a preliminary coding protocol with relevant themes based on the literature review, the research questions and the quantitative analysis. Examples of themes included in the coding protocol were: caregiver attitudes or feelings about the child´s school; caregivers’ expectations about preschool education; caregivers’ self-‐ efficacy in relation to their own life goals; caregivers’ views on the importance of observation for learning.
This protocol was then tested by this researcher through a preliminary analysis of the data via an iterative process, which included three rounds of coding with the data from five different cases. This researcher improved the coding protocol between each round and added some emerging themes. After these three rounds of coding, each time with an improved version of the coding protocol, the coding protocol seemed stable enough to proceed with the analysis of the rest of the cases. Appendix B shows the final coding protocol with the different hierarchies of codes and themes and each code/theme definition.
The researcher also trained the RA in the use of the N-‐Vivo software and the coding protocol, after testing and refining it, as explained above. The researcher and the RA then coded five more cases independently using the final coding protocol (in Appendix B) until an intercoding reliability of 0.80 was achieved. This was to enable the RA to code the data from the rest of the cases consistently with the coding protocol. This was the extent of RA involvement.
Once the data was coded according to the protocol, this researcher produced word documents with all the quotes that referred to each theme or category. Then, within each of these word documents (containing the quotes for a specific theme) this researcher grouped the quotes for each of the 30 children according to the HLLE level of the child´s home. So, for instance, within a specific N-‐Vivo–word document
containing all the quotes about a specific subtheme, the quotes of children from low HLLE were put together, as were those of the children from the mid and high HLLE categories. Subsequently, this researcher read each document several times in order to extract the main tendencies or trends for each topic. One of the aims of this first reading was for the local Chilean low SES HLLE themes to emerge.
This researcher noted the themes that emerged on the margins of the document. Then a short outline of the primary and secondary trends found in respect of each topic was made. Subsequent to this, the researcher went back to the literature to see how and whether these themes correlated with or had been identified in previous research. It was then sometimes necessary to go back to the N-‐Vivo document to see if the specific categories mentioned by the literature were indeed observable. Ultimately, the
researcher developed a quantitative checklist with aspects of the trends for a particular theme analysed. This quantitative checklist was necessary to control how
representative a certain qualitatively observed trend was amongst the sample and within each of the three HLLE subgroups. In this sense, it provided a control from concentrating on cases that were notable and tempting to focus on but not necessarily representative of the views of these families or specific HLLE subgroups within these families. Moreover, at this point in the analyses, this researcher was immersed in what Goldenberg et al. (2005) refer to as “steaming green Hell of context”… : thus the
quantitative checklist served as an intermediary step and mid-‐journey pit stop enabling this researcher to get some distance from the data before diving once again into its complexity. An example of this checklist is provided in Appendix F.
Via this iterative process, certain themes and subthemes, which seemed to capture and characterise the backbone of these Chilean children´s HLLEs, emerged and crystallized. A level of consensus around certain trends and themes emerged and around which specific cases and quotes could serve to illustrate these. This researcher then began to write up these thematic findings, which resulted in three chapters for the qualitative study, one focussing on general learning and parenting views, another one on HLLE practices, and a third chapter that explored these families´ beliefs in relation to
language and literacy development. . As is normal in these types of qualitative analyses, during the writing, new coding categories became clear. It was then necessary to go back to the N-‐Vivo documents to confirm whether a potential new subtheme was verifiable, in which case it was incorporated into the checklist. One example of a theme that emerged through this iterative process was that of “cossetting versus demanding too much from the child”. Quotes selected to illustrate the findings were translated into English by this researcher and then translated back into Spanish to ensure accuracy.
Two native Spanish-‐speaking educational researchers also fluent in English checked these translations and provided feedback and comments.