I'm trying to ensure the practical utility of the responses for informing practice is not lost through the process of raw data to codes then theme generation. To code and then create themes per question and then build overarching themes seems a good way to retain the meaning in responses when they move into codes then themes. If anyone know of an example of this they have seen in a paper I'd be grateful as I'm worried it's not allowed : ) Thanks in advance!
Per question you mean per interview question? In general, I code regardless of the questions - I try to track common code patterns in the data, with which I then build the themes. Once the themes are decently shaped, I can them think of a number of overarching themes. I never had issues in terms of retaining value for clinical practice using this method, so I'd definitely give it a try :)
I agree with Nicolò Zarotti that it often makes the most sense to code an entire interview for all the locations where a given code may occur. Brown and Clarke refer to working with a "corpus" of data, by which they mean the full data set, not just separate questions or separate interviews. In particular, if you limit your codes to what can be found in any one question, you may well miss connections between codes.
I also agree with Nicolò Zarotti and David L Morgan in focussing on the whole data corpus, and considering responses in turn, rather than by each question. You may miss the links participants make to previous responses if you split the data up by question. Best of luck with the analysis!
- Badges
- Science topic