,
10 tweets,
3 min read
Read on Twitter
1/10 I knew @ginnybraun & I couldn't tackle the use of saturation in #thematicanalysis in only 10 tweets! I want to come back to this paper & explain why most papers offering concrete guidance on saturation only work for coding reliability TA: journals.sagepub.com/doi/abs/10.117…
2/10 This & other papers use TA to analyse data & produce guidance on saturation - the problem is none define/locate their type or style of TA, explain the philosophical & procedural assumptions embedded within it, and how these diff from other types or styles of TA...
3/10 Guest et al. assume themes are like 'diamonds in the sand', entities that pre-exist the analysis, & so the analytic task is unearthing the themes that already exist in the data, we discuss this further in this commentary (on a statistical model): tandfonline.com/doi/abs/10.108…
4/10 So themes are conceptualising as 'analytic inputs', things that are identified at the start of the analytic process & guide the coding process. If your themes are your interview questions, we can start to see how it seems possible to judge saturation during data collection
5/10 Furthermore, themes are often conceptualised as what we call domain summaries (see an earlier thread on this), summaries of everything the participants said about a topic (classic theme titles are 'experiences of...', 'benefits of...', 'barriers to...' Again we can see how..
6/10 The notion of saturation makes sense if you're not looking for underlying patterns & concepts, but summarising surface meanings in relation to pre-identified topics; judging saturation during data analysis seems more plausible if these are your parametres
7/10 These papers often focus on 'code saturation', which often means identifying only *1* instance for a code to be 'saturated', again this reflects a very different conceptualisation of a code, something fixed & capturing relatively superficial meaning...
8/10 This paper (still problematic in some ways in terms of our type of TA) offers a useful distinction between code & meaning saturation, more conceptual codes are much slower to saturate than more descriptive codes : journals.sagepub.com/doi/abs/10.117…
9/10 So how many interviews are enough? For me & @ginnybraun saturation=information redundancy is too problematic a concept to use in our TA approach, the philosophical assumptions embedded within don't 'work' for our approach - too realist & doesn't gel with our analytic process
10/10 There are no easy answers to 'how many is enough?' Sorry! It depends on so many things - but pragmatic & practical concerns often win out... & we're okay with that, #thematicanalysis & qual research are messy, and that's okay. Qual research shouldn't be judged on quant stds
