Thematic Analysis

Braun and Clarke (2006), offer a complete and in-depth paper on what it is, guidelines to using it and pitfalls to avoid. They state that it can be considered a method in its own right - contrary to other authors stating it is not. It is compatible with constructionist paradigms and they stress its flexible nature in its use which can sometimes cause it to be framed by a realist/experimental method (though they don't particularly go along with this).
Importance is placed on explaining 'how' analysis is conducted (as its often omitted) as well as describing what and why they are dong it that way. Terminology is defined - Data corpus, set, item and extract. [all interviews, answers or themes, one interview and quotes].
" Thematic analysis is a method for identifying, analysing and reporting patterns (themes) within data." p.79
we shouldn't say that themes 'emerge' because actually that implies that they reside in the data - in actual fact they reside in the heads of the researcher who plays an active role in identifying, selecting and reporting them.(Ely et al., 1997 and Taylor & Ussher, 2001)
How do we say a theme exists? There are no hard and fast rules - though prevalence is important. A theme could be prevalent in one data item or across the whole corpus. And it may (or may not) be present in every data set or it may be present to only a small extent.
You should think about how a theme captures something of importance to the overall research question(s). This may make it 'Key'. The the question lies in how to report it ie 'most participants' or 'some..' or 'a number of...' etc.
You need to ask yourself whether you want to provide a rich thematic description of the whole data set or do you want to provide a detailed account of just one aspect of the data.
Next, will you provide an inductive analysis - whereby you link the themes to the data ie from specific questions, or a theoretical analysis - whereby your research questions evolve from the themes.
Semantic vs latent themes ie surface level descriptive or more deeper analysis of the underlying causes, assumptions and conceptualizations - this leads towards a more constructivist approach
Back to the paradigm wars - if a constructivist paradigm is used, then the themes are built from sociocultural contexts which enable individual accounts. In comparison the realism framework allows a more simple explanation to develop since meaning, experience and language are unidirectional. ie the language is used to describe experience and provide meaning.

The paper then goes on to describe 6 steps in thematic analysis - to be used flexibly and as a guide.
1. Familiarize yourself with your data: jot down notes for coding schemas that you will come back to in subsequent phases.
2. Generate initial codes. Work systematically through the data set  and identify interesting aspects in the data items that may form the basis of repeated patterns.
3. Search for themes: sort the potential codes into themes - broader analysis of whole data set. Use concept mapping
4. Review themes. this is a two step process - level 1 consists of looking at the extracts for each theme and deciding if they really fit that theme, and are coherent. if not, reassess the theme and perhaps discard extracts if necessary. Level 2 of this stage will depend on your theoretical approach and requires revisiting the whole data set and consider the validity of each of the themes.
5. Define and name the themes. Look at the data extracts for each theme and organise into a coherent account with an accompanying narrative. Look for sub-themes within a theme and finally, use clear (short) names for the themes so that the reader understands quickly what it means.
6. Produce the report. More than just describe the data, tell the story by using the extracts to make an argument in relation to the research questions.

Some questions to ask towards the end fo the analysis:
what does this theme mean?
What are the assumptions underpinning it?
what are the implications of this theme?
what conditions are likely to have given rise to it?
why do people talk about this thing in this particular way?
what is the overall story the different themes reveal about the topic?

Potential pitfalls are described:
1. no real analysis is done and the analysis is just a sting of extracts with little analytic narrative.
2. uses the data collection questions as the themes.
3. no coherence around an idea or concept in all aspects of the theme.
4. no consideration of alternative explanations or variations of the data
5. mismatch between the interpretation of the data and the theoretical framework.