CHAPTER 3 METHODOLOGY
3.3 Data analysis
As is typical of qualitative research, data analysis in this multiple-case study was on- going, recursive, inductive and data driven (Duff, 2008; Taylor & Bogdan, 1998). Throughout the data collection period, I read my notes and the written documents multiple times. I also listened to each interview several times while preparing follow-up questions or transcribing. In particular, each time I drove the almost hour and a half between my house and Hope College I tried to use the time to listen to the recordings of previous interviews. As Duff (2008) points out, “from the earliest data collection and transcription stages, […] data analysis is already taking place” (p. 159).
After I had transcribed most of the interviews, I started the coding process. Following Mackey and Gass (2005, p. 241), the data was initially analyzed through a process of “open coding” in which I looked for “anything pertinent to the research question or problem, also
bearing in mind that new insights and observations that are not derived from the research question or literature review may be important.” Because I was not sure what would turn out to
be relevant for the way my participants‟ negotiated literacy practices in college, and because I thought it would take me too much time to decide whether each topic in the interviews might eventually be important or not, I decided to code all the lines of all the interviews. This coding process was extremely time-consuming and quite tedious, but, after it was done, it made the enormous amount of interview data much more manageable and, most importantly, easily searchable. I used the qualitative data analysis software Atlas.ti to organize my coding. Even though learning to use the software required a significant time investment, it was certainly a very
useful tool for organizing data, coding and storing memos. As Séror (2005) claims, this type of software can “facilitate the mechanical steps in the process of analysis” (p. 323). Even though
Atlas.ti has many different possibilities for data analysis, and I did spend some time trying to learn those to see how useful they could be, I ended up deciding not to use these more
sophisticated features because I thought the learning curve was not worth it, and I could not really see any added benefits in using them. I essentially used Atlas.ti for coding and memoing. It may be worth mentioning that I did all the coding and memoing myself, just using the software to record and keep track of them. In other words, I did not do any automatic coding, which I thought would not be adequate for the kind of qualitative analysis I was engaged in doing.
Because I used open coding, I ended up with almost 200 codes. Clearly, if it were not for the software, I would not have been able to keep track of these. At the same time, if it were not for the software, I would probably have been more conservative coming up with different codes. I used codes for different purposes. I had several codes that referred to what the speakers were saying. Examples of codes in this category that occurred frequently are:
anxiety/insecurity/stress/tired asking for clarification/help (or not) classroom practices
comments on professor confused/unprepared
explanation for poor performance grades
interactions with professor (or not) note taking/highlighting
professor‟s feedback
resorting to others study guide
tests/exams/quizzes working with tutor/mentor writing assignment
writing center writing process
I had different codes for each of the courses that required a significant amount of reading and writing such as ENG101, ANT200, GOV211 and SOC200. I had a code for each of the participants, which was used each time a participant was mentioned by another participant or by a professor. I also had codes to identify what I was doing in the interviews. Examples of these are: requesting material (which was used 202 times!), researcher clarification, and research procedures. I also had separate codes for the faculty interviews such as faculty: comments on participants, faculty: reading, and faculty: writing.
For each section of an interview, usually more than one code was assigned. For the section reproduced in Figure 3.1, for example, the following codes were assigned:
ECO 110
grades: expectations vs actual multiple choice
strengths/self-confidence tests/exams/quizzes weakness/difficulty
E: Huh. Were you surprised by the grade?
So: Super surprised cause I felt like I did good. But on the essay writing part, I did good, like cause the essay writing part was 14 points and I got 12 in one of them and I got 11. And he said like to get 14 is really difficulty. You have like 12 everything. So, in the writing part I did good, but the problem is,
E: So it was the multiple choice, the essay and what was the other part? So: Writing short answers.
E: Short answers, okay. And what did you get in that one?
So: I did good, cause that‟s where my strength is cause I like writing down what I think instead of looking and, I hate multiple choice, I just, I never do good in multiple choice, it‟s
something I have to really work on hard.
E: Especially because all the tests are going to be like that, right?
So: Yeah. And his questions, oh my God, he will twist them around and it‟s like the same thing, but just, this was ridiculous.
Figure 3.1 Extract of interview
As analysis progressed, I started “examining the data for emergent patterns and themes” (Mackey & Gass, 2005, p. 241) by recursively going through the data. The search options
available in Atlas.ti made the process of revisiting the data somewhat more focused and efficient. For example, when I started to look for patterns in reading challenges, I went back to the
interviews searching for all instances of the following codes: reading difficulty, reading load, reading process, and reading to write.
Interviews were a major source of data in this investigation, and they gave me important insights into the participants‟ perceptions of their literacy experiences (an emic perspective).
Other themes, however, emerged mainly through the analysis of the written documents. These themes were often different from the ones voiced by my participants. I believe that by reporting on both types of themes, I can better address the research questions. As Leki (2007) explains:
In analyzing the data, rather than detailing only the research participants‟ emic
perspectives and limiting this report to the themes that were the most salient for them, I elected to also elaborate certain themes that will be, I believe, of most use and interest to writing teacher/researchers. (p. 8)
Each of the seven case studies were first analyzed individually and then a comparative cross-case analysis was done (Duff, 2008). Whereas the individual cases yielded insights into the particular experiences each participant went through, the comparative cross-case analysis pointed to “important theoretical questions and connections” (Lea & Street, 1998) regarding the literacy
experiences refugee students go through in their first year of college.