Qualitaive Data Analysis Protocol
This document is still a work in progress.
Data Collection
Most of the data subject to QDA comprise transcribed interviews. Interviewees are selected based on a theoretical sampling strategy, which is responsive to prior analysis. Interview guides are informed by case files, which include detailed profiles of prospective interviewees, which gauge the value that they will bring to the research project as a whole. These guides are in turn informed by memos written during the coding process and in response to situational mapping excercises.
Data Processing
I use an entirely locally-hosted whisper model to generate an initial transcript of the audio recording. I then manually edit the transcript to correct any errors in the computer-generated version. Given the accuracy of the whisper model, corrections tend to be simply resolve mis-spellings of proper names of people and projects (and their acronyms), re-assign my affirmations to interviewees’ responses to my own speaker label, and correct any minor issues deriving from the model not properly interpreting interviewees’ accents. I then run a basic spell checker to find any blatant errors and to ensure consistency. I rely on find and replace functions to resolve these issues en masse.
Some changes to the transcript can not really be anticipated during the data preparation phase. For instance, during coding, I may modify the transcript to make minor correctuons that I missed while transcribing, such as mis-capitalization or unresolved trailing blank spaces following a terminal period. These corrections are fine as long as meaning is preserved or clarified, and as long as the trancripts and qc coding files remain consistent. Regarding this latter point, since the reoslution is limited to line numbers, this will be ok as long as I add an equal number of correspnding lines to the same position in the codings file.
Once the project is complete or a transcript is fully coded, I will copy the file from the qc corpus directory into the data directory and mark it up using markdown for easier publishing.
Coding
See my notes on coding and my notes on specific coding methods.
Open coding
Open coding (also sometimes referred to as initial coding) constitutes a first pass through data. I go through the transcript line by line, remaining open to all theoretical directions and interpretations that arise. It is akin to a brainstorming session.
Essentially serves asn opportunity to reflect deeply on the contents and nuances of the data and to begin taking ownership over them.
I also engage in open-ended memo-writing pertaining to specific passages within a transcipt, similar to writing in-document memos using MaxQDA. These are stored in a single file associated with the interview transcript, with reference to specific line numbers and labelled subheadings when appropriate.
This also relates to holistic coding, which is similar to open/initial coding, but operates at a coarser grain. Holistic coding is useful for “chunking” the data into broader topics as a preliminary step for more detailed analysis later on. However, for the sake of simplicity, I not really make this distinction between open and holisitc coding methods in my coding processes.
- Prefix:
O: - Notes on open coding
In vivo coding
In vivo coding draws out specific words or phrases directly from the text. In vivo codes are identified opportunistically, and are typically created during an initial or open coding phase.
In vivo codes are especially useful for drawing attention to the improvised collective adoption of certain terms or imagery.
- Prefix:
IV: - Notes on in vivo coding
Process coding
Process coding entails framing expressions of action in terms of generalizable processes. This may help reveal common objectives, circumstances and tooling that contextualize these action or that give these actions meaning. It may also help to elucidate commonalities and differences across similar processes, especially when coding the same passages with multiple processes.
- Prefix:
PRO: - Notes on process coding
Values coding
Entails three kinds of codes:
- Values
- The importance that respondents attribute to themselves, to other people, things or ideas.
- The principles, moral codes, and situational norms that they live by.
- Attitudes
- The way respondents think and feel about themselves, other people, things or ideas.
- They are part of a relatively enduring system of evaluative, affective reactions.
- Beliefs
- Part of a system that includes respondents’ values and attitudes, plus their personal knowledge, experiences, opinions, prejudices, morals, and other perceptions of the social world.
- RTey can be considered as “rules for action”.
Some key phrases that signpost values, attitudes and beliefs include:
- It’s important that…
- I like…
- I love…
- I need…
- I think…
- I feel…
- I want…
- Prefix:
VAL:for preliminary or general codings;VAL-V:,VAL-A:orVAL-B:when specifying a specific kind of value code. - Notes on values coding
Versus coding
- Prefix:
VER: - Notes on versus coding
Domain / taxonomic coding
- Prefix:
X: - Notes on domain and taxonomic coding
Attribute coding
Maintain a separate document with poeple’s and projects’ names, affiliations, objectives, etc, as a way to keep track of them all. This may turn into a memo document in its own right.
Memos
See my notes on memo-writing methods.
Kinds of memos
I find memos hard to classify and distinguish. For instance, when I attempt to simply describe an interaction that I observed or participated in, which would typically be classified as descriptive field notes, I often find myself offering interpreations alongside my accounts of what happened. At the same time, there are some clear distinctions between these accounts of events, memos about situational maps, or general reflections about the strange epidemiological world I’m encountering. I therefore lump most of these into the category of “memo” (and stored under the data/memos/ directory), and append tags to the metadata that indicate the context in which the memo was created. Some common tags include:
- situational map
- in-document memo
- descriptive field notes
- post-interview reflections
- reflection on an encounter
See Mruck and Mey (2019: 483-485) for further details on the distinctions and crossovers between journalling and memo-writing.
Titles and descriptions
I find it kind of difficult to provide very brief titles for memos, but there is a lot of value in writing brief descriptive summaries. Those are probably better quality and more helpful than the titles themselves.
I also struggle with the file names. Sometimes it is appropriate to use a very simple series of hyphenated words, especially if I plan on adding content or editing the document in the future. But in situations where I am reflecting on a speciifc encounter, I may prefer to use a datestamp, with or without a series of descriptive terms (usually borrowed from the title).
For memos that are linked to other documents, such as in-document memos associated with a collection of files collected during an interview (transcript, recordings, etc), then I put the memo in the folder where those materials are kept and five the memos corresponding names.
Analyzing memos
I may import memos into qc for their analysis as research materials in their own right. I prefer to analyze memos that are more descriptive and less interpretive, or that have a basis in what my informants say, rather than what I say.
Diagramming
Birks and Mills (2015: 100) claim that “diagramming is the creative tool to use when operationalizing the logic of abduction”. (gorra?) expands on this by stating that it enables seeing the data in new ways and enabling flexible movement