Thematic Analysis

Data terms

data corpus - all data collected

data set - all data from corpus being used for specific analysis

data item - an individual piece of data from the data set

data extract - individual coded chunk identified within data item

types of thematic analysis

framework approach

empirical phenomenology

content analysis

grounded theory

hermeneutic approaches

interpretive phenomenological

consensual qualitative research

generic thematic analysis

empirical phenomenology

early form of thematic analysis

focuses on in-depth analysis, often of single cases at first

goal is detailed description of features and variations in phenomenon of interest

Hermeneutic approaches

more flexible approach

argue for going beyond basic protocols to look at implicit/unconscious meaning within texts

interpretation and understanding

emphasizes role as interpreter

interpretive phenomenological analysis

accessible, systematic, practical approach to dealing with phenomenological data

includes specific steps to follow (how to develop higher- and lower-order categories within data)

places where there are similarities or differences in experiences

content analysis

balances between quantitative and qualitative methods

qualitative data (open-ended responses) > quantitative output (frequencies of specific categories/codes)

in modern data analysis, we will often use computerized methods to do this - linguistic inquiry and word count (LIWC)

word clouds as exploratory qualitative analysis

NVivo

framework approach

structured approach using framework matrices

central features: detailed coding framework a-priori based on theory, previous research, questions asked, ect. some inductive coding during data analyses

process is shown with diagrams, making coding process more transparent

• started to manage policy research

grounded theory

a way to analyze data and generate theory

data collection and analysis are concurrent, until “theoretical saturation” is reaches

low level abstraction > abstract theoretical concepts

identify codes (irritated, hostile, frustrated, sad, fatigued, suicidal) > identify concepts (anger and sadness) > identify broad categories (depression) > develop theory (two components of the experience of depression)

specific purpose of generating theory, keep recruiting people until you are no longer getting new information

consensual qualitative research

distinguishable by its use of multiple analysts and auditors > seeks to find consensus among multiple researchers looking at the same data

primarily social constructionist, with some post-positivist learnings in terms of reliability of measurement

Uses frequency in data to identify

general themes (seeking consensus about core parts)

• reported by all or almost all participants

• may define the experience

typical themes

• reported by at least half of the participants

• useful for constructing narrative of typical experiences

generic thematic analysis

thematic analysis is a method for identifying, analyzing and reporting patterns (themes) within data. It minimally organizes and describes your data set in (rich) detail.

generic thematic analysis > summary of the process

raw data > interview transcripts

codes > drawn from raw data, often 100s

themes (codes grouped into broad topics)

generic thematic analysis - history

prior to braun & clarke (2006) > there was no well-defined approach to “thematic anaylsis”

other techniqies available > but many ascribe to a specific philosophy (phenomenological)

benefit > can be used without being tied to specific philosophy so can be used more widely than some qualitative methods

initial considerations

choosing your sample

type of thematic analysis

types of themes

epistemology

choosing your sample

the sample has to be “theoretically interesting” > a convenience sample is inappropriate

only “exceptional” people > exemplify something of particular interest

> all have clinical depression

> all in the top 10% on perfectionism scores (known as “intensity sampling”)

ideally, matched on core demographics and other confounding variables

types of thematic analysis

inductive > being analysis without preconceptions and simply describe what you find

theoretical > pay special attention to particular themes in the data which you decide on beforehand

semantic themes

make no inferences; do not look beyond what the participant has said or written

note- does not mean no interpretation or theorizing, but that it is done based on patterns within explicit / surface level meanings

epistemology

essentialist / realist > reports the experiences, meanings and the subjective reality of participants, most common in clinical psychology

Constructionist > focuses on how things (events, realities, meanings, effects of a range of discourse operating within society. tend to focus on latent themes

how to conduct thematic analysis

1. familiarize yourself with the data

2. generate initial codes

3. search for themes

4. review themes

5. define and name themes

6. produce report

familiarize yourself with the data

read and re-read all of your raw data, taking notes as you go

one good way to start this is to do your own interviews and / or transcribe all your own data

generate initial codes

decide on the size of the data items (split transcript into equal parts) - each item covers an equal unit of meaning. This is frequently, but not always a roughly equal amount of text

assign a short code which summarizes the content of each data item equally - codes must use language not numbers

search for themes

begin to sort the codes into similar groups. Those groups are called “themes”

ways to do this - highlighter, post-it notes, visual diagrams, qualitative analysis software

often give themes working names, but at this early stage it’s still a work in progress

review themes

cross-check your codes with themes- can your themes account for all you codes?

probably not > you may need to change some of your codes and/or some themes

ideally, your themes would cover all of the data > in practice, will probably have one “garbage bin” theme those don’t fit anywhere

define and name themes

solidify final themes > provide short definition and name that summarizes the data concisely and completely

you can have subthemes within each larger theme that are worth discussing

at this point, others should be able to use these definitions to look for these sorts of themes in another dataset > you will have created a codebook, of sorts

produce report

start writing the paper > the writing IS the analysis

craft a detailed argument of the themes and use compelling extracts to support your points

relate your findings back to prior literature and your research question

a quantitative approach to qualitative data

once the themes have been identified, they can be converted to quantitative codes for use in analysis

often dichotomous 1 = present 0 = absent

other options - coding frequency, coding on an ordinal scale if the intensity of the theme was important

analysis tips

themes converted into numeric codes > analyze like any other quantitative variable

inter-rater reliability is crucial

• Kappa (categorical)

• Intraclass correlation (interval or ordinal)

• rule of thumb > = .70 is acceptable

tests of validity are also helpful

• convergent validity: do your themes of intimacy correlate with a questionnaire measure of intimacy

evaluating analysis

data have been interpreted, not just described

extracts support claims

analysis is convincing, well-written and organized

enough time has been allocated to analysis

written report

assumptions are clearly explicated

consistent epistemological position

researcher is active; themes don’t “emerge”

not acknowledging active role (identified theme, it didn’t emerge)

pros of thematic analysis

make sense of a lot of data

accessible

very flexible

relatively easy/ quick to learn

accessible to wide audience

usefully summarize large body of data and or offer detailed description

can generate unanticipated insights

cons of thematic analysis

time consuming

leave room for error

hard to do larger studies

drift in coding

confirmation bias

people often fail to follow strict guidelines, leads to poor analysis

cannot move beyond description unless within a theoretical framework

misses a lot of nuance compared to other qualitative approaches

cannot determine cause and effect