We began the QDA (qualitative data analysis) part following on from the interview with Alice Podd, in search for themes and insights. Abbey has pawed a detailed report of the insightful points made in the interview transcript. Next step is to code the data in search for themes, like behaviours and specific acts or relationships and interaction. Beforehand, I want to research coding qualitative data so that detailed data analysis is achieved and the data is organised into specific themes or topics.
How and What to Code
Coding is the process of marking passages of text that are about the same thing, say the same thing or discuss things in the same way. Codes can be based on:
- Themes, topics
- Ideas, concepts
- Terms, phrases
A Priori codes can be identified from a range of sources:
- Previous research or theory
- research or evaluation of the questions you are addressing
- Question topics from interview schedule
- Gutt feeling about the data
So in the case of the interview with Podd, our priori codes could be:
- Aims of Amnesty
- Major campaigns
- Pain points
- Types of demos
- User groups
- Contact details
Grounded codes emerge from data because you put aside your prejudices and previous knowledge of the subject area and concentrated instead on finding new themes in the data. Grounded codes are particularly insightful because they do not contain bias presumptions, they are data based and have higher credibibility in that sense. Of course, grounded codes emerge from the interview data and cannot be thought of before the initial data analysation.
What to look for when writing in the grounded theory tradition, Charmaz (2003: 94-95) suggests you ask:
- What is going on?
- What are people doing?
- What is the person saying?
- What do these actions and statements take for granted?
- How do structure and context serve to support, maintain, impede, or change these actions and statements?
A more detailed list of the kind of things that can be coded can be seen in Table 1 below.
There are many suggestions about the approaches to data in so that you remain open minded about what can be coded and begin to notice significant patterns in data. The most famous approaches are those made by the grounded theorists.
A common procedure recommended by theorists is ‘constant comparison’, meaning every time you select a passage of text (or its equivalent) and code it, you should compare it with all those passages already coded, perhaps in other cases. This ensures that your coding is consistent and allows you to consider the possibility either that some of the passages coded that way don’t fit as well (and might therefore be better coded as something else) or that there are dimensions or phenomena in the passages that might well be coded differently. The potential does not stop there as passages can be compared with those coded in a similar or related way, even compare them with cases and examples from outside your data (Gibbs, 2017).
For example, Strauss and Corbin (1990) devote a chapter called ‘Techniques for Enhancing Theoretical Sensitivity’ to examining some of the ways they recommend qualitative analysts might use to ensure that they look carefully at the data and explore all its dimensions (Strauss and Corbin, 1990 pp.75-95). This includes systematic comparisons and far out comparisons.
In the former, you think about all the ways in which some phenomenon you have found in the data can vary and be treated and seen differently by people. For example, if your respondent has been talking about the way her parents continued to give her financial support after she had left them and set up her own home, you can compare this with all the other ways that parents might support their children, financial, emotional, child care, finding employment, setting up personal contacts, housework, do-it-yourself, gardening and many others. This allows you to think or discover new ways of coding the experiences this respondent and others in your study might have had.
In the case of far out comparisons, the comparison is made with cases and situations that are similar in some respects but quite different in others and may be completely outside the study. For example, still thinking about parental help, we might make a comparison with the way coaches help sports men and women. Reflecting on the similarities and differences between coaching and parental relationships might suggest other dimensions to parental help, like the way that coaches get paid for their work but parents don’t.
Other techniques to identify themes and codes
Ryan and Bernard in a recent paper (2003) suggest a to code transcripts to disorder themes in data. They draw heavily on Strauss and Corbin (1990), suggesting:
- Word repetitions – look for commonly used words and words whose close repetition may indicate emotions.
- Indigenous categories (grounded theorists refer to this as in vivo codes) – terms used with a particular meaning and significance in their setting.
- Keywords in context – the range of uses in key terms in the phrases and sentences in which they occur.
- Compare and contrast – essentially the grounded theory idea of constant comparison. Ask ‘what is this about?’ and ‘how does it differ from the preceding or following statements?’.
- Social science queries – introduce social sciences explanations and theories, for example, to explain the conditions, actions, interaction and consequences of phenomena.
- Metaphors and analogies – metaphor to indicate something about their key and central beliefs about things, these may indicate the way they feel about things too.
- Transitions – one of the discursive elements in speech which includes turn-taking in conversation as well as the more poetic and narrative use of story structures.
- Connectors – connections between terms such as causal (‘since’, ‘because’, ‘as’ etc) or logical (‘implies’, ‘means’, ‘is one of’ etc).
- Unmarked text – examine the text that has not been coded as a theme or even at all.
- Pawing (i.e. handling) – marking and scanning the text. Circle words, underline, use coloured highlighters, run coloured lines down the margins to indicate different meanings and coding. Then look for patterns and significance.
- Cutting and sorting – traditional technique of cutting up transcripts and collecting all those coded the same way into piles, envelopes or folders or pasting them onto cards. Laying out all these scraps and rereading them, together, is an essential part of the process of analysis.
Types of coding
When beginning to code data, the tendency is to create codes that are some kind of summary of the text you are examining. This process, namely descriptive coding, essentially forms a summary description of what is in the transcript or text. An essential part of QDA is the development of codes that go beyond the description and start to categorise and analyse the data.
One example of descriptive coding might be that the analyst reads the text carefully and circles what seems to be key terms, events or actions. A short note of what these are could be written besides the circling. This is the start of descriptive, or what grounded theorists refer to as open coding. The analyst would then illustrate an initial coding list housing terms which would summarise the events/actions noted by the initial coding.
In another example, the analyst might make use of a wide margin on the right of the transcript, so that code labels and other comments can be written in the space. When printing the transcript from the interview with Podd, we made sure that there was space on the right to allow for comments. The analyst might use brackets on the right of the transcribed text, coding much larger passages or chunks of text. This form of coding is most useful when making retrieval i.e. gathering together all the text about one topic. With larger chunks, the retrieved text is less likely to be decontextualised. The analyst would use a highlighter to identify words that refer to feelings. See Example 2 below:
Organising the code into a list or coding frame:
As well as marking the transcript or field notes to show what is coded as what, a separate list of the codes constructed should be kept, as well as a short definition against each one. Next time a passage could be coded as an existing code, refer to the list or frame to see if it exists and if it does, check the definition to be sure that it does fit in. If an appropriate code can’t be found, then a new code can be created.
A large number of codes will eventually exist, which can then be sorted into groups or some sort of order. A method used to order lists is hierarchy, whereby several code groups can be found based on their types or kinds of something. In that case, they can be placed together and housed in a list of their own, or sub-codes.
Two thighs may emerge: Firstly, could start categorising codes, forming the bases for a between cases comparison or analysis. Secondly, new possibilities suggest themselves when looking at the groups of similar codes. This is what Strauss and Cobin (1990) refer to as ‘dimensionalising‘. People do, react, categorise or cause things in a number of different ways. These different ways are referred to as dimensions of a thing, according to Strauss and Cobin. For example, people tend to want to take a break from work. They may do this in an enormous variety of ways – they may take a holiday, go for a drink, go for a walk, take drugs, meet up with friends etc. These are all dimensions of ‘taking a break’.
Several things are suggested by such dimensions. First, dimensions may be thought of that were not initially coded, which can then be included in the coding list and if not, an explanation as to why. Second, dimensionalising and categorising will begin to raise questions about the relationships between codes (do those who have been coded using one particular code tend also to be coded in another particular way?) or between cases (why are these cases coded this way and other cases in a different way?). Thus, this kind of development of coding and reorganisation of codes can form the basis for some key analsis of the data.
Non-Hierarchical coding (flat coding)
A non-hierarchical arrangement of codes, like a list, there are no sub-code levels. Example:
- Close, generalised friendships
- Sporting friendships
- Sports club members
- Work friends
- Making new friends – same sex
- Making new friends – different sex
- Losing touch with friends
- Becoming sexual relationships
Hierarchical coding (tree coding)
A hierarchical arrangement of codes, like a tree, there are a branching arrangement of sub-codes. Ideally, codes in a tree relate to their parents be being ‘examples of …’, or ‘contexts for …’, or ’causes of …’, or ‘settings for …’ and so on. Example:
- Friendship types
- Close, generalized
- Changes in Friendship
- Making new friends
- New same sex friends
- New different sex friends
- Losing touch
- Becoming sexual relationship
- Making new friends
Applying new codes
New codes are likely to be created during the data coding process. The analyst would need to go back and check the units of data coded previously, in order to check if there is any more data that should be coded at the newly created note.
Memos and codes
It is important that the analyst keeps meaningful written notes made during the coding process, often called memos. A major use for memos is to record longer definitions of the codes and to note any analytic thoughts made about the significance and relationship to others of the code in question.
Typically, a memo should incorporate:
- Why the code was created
- Detail of what the code is about and what the coded text reveals
- Why the code has been changed (i.e. renamed)
- Thoughts and questions about the analysis that occurred
To summarise …
- Coding involves categorising and indexing sections or chunks of data.
- Codes can come from theory and explanations outside the data (A priori codes) and/or emerge from the data (grounded codes).
- Data formats that can be coded range from transcribed text to video.
- Coding often starts by being descriptive, then develops to being analytical.
- Any new codes created should be applied to the whole data set (previously coded units of data).
- Memos should be used to record thoughts and ideas about your codes during process.