• No results found

3.3 Data Collection

3.3.2 Visual Materials

3.5.1.1 Thematic Analysis’ Phases

The initial phase for analyzing data begins with the researcher becoming familiar with the data to the fullest extent. To do so, researchers must read through the entire data set (Braun & Clarke, 2006), especially if they were not a part of the data collection process. It is advised that the researcher takes notes throughout, as they will be used in later stages. Accordingly, I audio recorded all of the interviews and had them professionally transcribed. To begin, I read all of the interviews without making note of anything. I simply wanted to immerse myself in the data and

become even more familiar with the participants’ stories. I then re-read all of the transcripts and began to make note of codes and ideas, aligning with the next phase of analysis. In the second stage, the researcher needs to have the ability to sense a codable moment (Boyatzis, 1998). It is in this manner that themes begin to emerge. In order to get there, however, there needs to be a set of codes first, as codes help inform and formulate a list of potential themes. These codes are usually a result of things that appear interesting to the researcher and that are connected to the larger topic of the given study. It is important to note that codes are different than themes; they tend to be smaller and more specific to the data (Braun & Clarke, 2006).

Since I was interested in discussing particular types of vulnerabilities and protective factors related to the participants’ social identities, I was open to codes and ideas that emerged from the questions in the interview protocol. However, precisely because I knew that I wanted to reveal a different set of vulnerabilities, I also focused on codes that were particularly related to the six social identities discussed here. As such, I engaged in both inductive and deductive coding (Merriam, 2009). Since I wrote analytic memos (Merriam, 2009; Saldaña, 2013) for each of the interviews, I also revisited those during the coding phases. These memos were related to resilience, social identities, participants’ behavior, and things to follow-up on. Memos were especially helpful as I prepared for the second interview.

The third stage is reached when there is a list of codes that has been defined. The codes are then sorted into potential themes with each of the codes categorized under each of them (Braun & Clarke, 2006). Braun and Clarke (2006) state that in this stage “you are starting to analyze your codes, and consider how different codes may combine to form an overarching theme” (p. 19). As this stage is undertaken, it is useful to organize the data using some sort of table or map to help the researcher visualize the data in its entirety. As such, once I developed a

solid list of codes on a word document, I defined each of them in order to have a level of consistency by the time I engaged in line-by-line coding. I then arranged the codes into categories (Saldaña, 2013), allowing me to place the already existing codes within larger categorical concepts. Codes were grouped together based on similarities to one another but always in relation to the larger concept that they were under. Once codes were grouped together, I deleted those that seemed duplicative.

Upon finalizing the codebook (Saldaña, 2013), I uploaded the transcripts and list of codes with definitions to NVivo, which is a qualitative data analysis software. Once everything was uploaded, I began coding the transcripts. On NVivo, I engaged in line by line coding by applying a code to specific sections of the transcript. Initially, I only coded some of the transcripts because I wanted to test the accuracy and completeness of the codebook. After this initial phase of coding, I made changes to the codebook. Specifically, I rearranged some codes by moving them from one category to another and I also added others. Once transcripts were coded, I was able to separate data through different filtering on the software. For instance, I was able to separate data based on codes that were specific to the research questions. I was also able to run single and double code queries. This then allowed me to develop a list of vulnerabilities and protective factors as potential themes based on analysis of the codes (Braun & Clarke, 2006).

By the start of the fourth phase, a list of potential themes has been outlined. Such themes are then analyzed in-depth to figure out whether there is enough data to support them (Braun & Clarke, 2006). As I solidified the list of potential themes, I separated the codes that they were most aligned with on NVivo to ensure that there was enough data to support each of them. Because I knew that I was going to discuss vulnerabilities and protective factors, I divided my themes into those two overarching categories. At the same time, I individually separated codes

into the six identities discussed here, which made it easier for me to situate them within a resilience framework. By the fifth stage, I had a definite list of the themes (Braun & Clarke, 2006). Due to the intersectionality of participants’ identities and the prevalence of systems of oppression, there was a significant amount of overlap in the themes. According to Wertz, Charmaz, McMullen, Josselson, Anderson, and McSpadden (2011), the intertwinement in themes is a sign of good work, as categories that are too separate from one another appear artificial.