Thematic Analysis: Qualitative Research Methods PDF

SPEAKER 0 Yes. Resume. Okay. So last week I gave an overview of qualitative research. What this is what the assumptions are. And we focussed on two specific methods of collecting data. So I focussed on interviews and I focussed on focus groups as well. Today we're going to talk about a method of an...

SPEAKER 0 Yes. Resume. Okay. So last week I gave an overview of qualitative research. What this is what the assumptions are. And we focussed on two specific methods of collecting data. So I focussed on interviews and I focussed on focus groups as well. Today we're going to talk about a method of analysing data a broad range of data. And that's thematic analysis. So some learning outcomes for today. By the end of this lecture followed by some reading and some practice. Uh, we will explain the purpose of thematic analysis. When might you contact thematic analysis on the data that you have and also to describe the methodological procedures associated with thematic analyses. So I brought some examples with me. And we will be doing an activity as well, um, in a few slides just to grasp the concepts associated with the magic analysis. So that's what you need to know with the assessment. But as I mentioned last week as well, thinking beyond the assessment of this module, for you to be able to then go away and conduct thematic analysis on the data that you have. Okay. So let's begin by first talking about what do we mean by thematic analysis. So thematic analysis is a method in qualitative research that is used to identify, to analyse and to report patterns of meaning or themes across our data set. Now we're looking at recurring ideas across our data set and ideas that are relevant to our particular research question. So our research question will guide our analyses. Why do we teach thematic analyses as part of our undergraduate and postgraduate courses? It's quite a foundational method that underpins many other approaches in qualitative research. So some of the skills that you're developing, for example coding, for example, organising those into broader ideas or themes and analysing those. These are useful skills for other types of qualitative analyses as well. And it's also quite a flexible approach. We can use thematic analyses with quite a wide range of data with primary sources, interviews, Abuse focus groups with secondary sources as well, which we talked about last week. And it's particularly suited for most types of research questions. And also it can be used with various epistemological standpoints. So thematic analysis is not tied to any pre-existing theoretical framework about knowledge about meaning about data. So for example you could be taking an essentialist or a realist approach to your research. You might be interested in investigating people's experiences, the meanings of those experiences and what you identify from. For example, your interviews or your focus groups is people's words provide direct access to the inner world. So if I want to investigate what are people's experiences of learning statistics in higher education, I go and talk to my participants. So if I adopt an essential is a realist approach. The things that people tell me in the focus group in the interview, those reflect their perceptions, their experiences, the meanings they associate with them. We can use thematic analysis from a critical realist approach as well. For example, this is where the focus is how individuals make meaning of their experiences influenced or based on their sociocultural situation. So what people tell us that provides access to their particular version of reality. So it's quite a flexible approach in that sense. But What's really important. Any researcher doing thematic analyses. We need to make a series of decisions, and we have to be very clear about those decisions, particularly when we report our analyses. So we need to decide our research question, our theoretical assumptions associated with our research questions, and the type of thematic analysis that we will be conducting. So I'm presenting an example here from Malik and Coulson's paper. Quite a nice paper in terms of how they presented their thematic analyses, how they've presented and talked about and interpreted the findings. So, for example, the present analysis was conducted in an essential realist frameworks where themes were developed or identified at the semantic level. So we're being very clear here on what approach we're Were taking. So I talked about thematic analyses as a method for developing for identifying patterns of meaning or themes. So what is a theme and what constitutes a theme? So a theme is a patterned response in your data set that has meaning in relation to your research question. I'm going to show you an example from published literature. So I chose my paper that I published recently with a colleague in the School of Psychology. Sorry, this is a lot of scrolling. It was a big paper, a lovely paper. So a few years back with Laura, we wanted to investigate people's perceptions and experiences with Covid 19 and decision making. At a time where the UK entered a phase of living safely with Covid 19. So we conducted a qualitative questionnaire. We analyse the data by means of deductive thematic analysis. And I'll come on to that. And we've developed themes that help us explain what are people's experiences with Covid 19 at this phase in the pandemic. So we've identified one overarching theme transition back to normal routines with four broad ideas illustrating people's perceptions and experiences. Living with the uncertainty. Theme number one. Concern for others. Theme number two the multiple consequences of Covid 19. Sense of control. Theme number four. And then we went on to talk about our themes and the ideas within those themes with supporting evidence. So this is what we mean broadly by themes in thematic analyses. So. What counts as a theme and what size does the theme need to be evident in within our data set, both across our data set and within a single data item? Now, data set is the data that you're using in your analyses. If you've come if you've contacted ten interviews, those ten interviews are your data set that you will be analysing. One in interview is one data item from that data set. Now, in working out what patterns are relevant and important in relation to your research question? Yes, it can be a question of prevalence, but also is more than that. It's about something capturing something important and meaningful towards your data. In relation to your question. Now numerous patterns could be identified across your dataset, but it's up to us as researchers to identify and present the ones that are relevant and meaningful. I'll give you an example. If somebody were to conduct some research looking at workplace satisfaction, well-being and so forth, might contact some interviews and patterns might be developing. I might be noticing that people are talking about the time that they arrive at work, or I try and be there by about half past half an hour before I start working. Now people arriving at 8:30 in the morning. We need to think, how does that relate to my research question? What does that tell me in relation to workplace satisfaction or wellbeing? Probably not much, but if I look at another pattern where people tell me that they try and arrive, give themselves at least 15 20 minutes to have those conversations in the common room, to have a cup of tea in the morning, to have a corridor conversation. Good morning. It helps with building interpersonal relationships in the workplace. That particular pattern? Yes, that's relevant to my research question. So we have to make a judgement as researchers in relation to the patterns or themes that were developing. Now, an answer to the question of what proportion of your data need to be need to display evidence of the theme? Ideally, there will be a number of instances of evidence of the theme across your data set, but it could be the case that one particular data item or a number of data items might not show evidence of that theme. And then it's going back to prevalence, but also what's meaningful and important towards your research question. Okay. So let's think about the types of thematic analyses. Now our themes, the patterns that we're developing within the data. And you'll notice the language that I'm using I'm not talking about themes emerging themes appearing and so forth. This is a very active process. As a researcher, I'm not going in my interviews and looking for themes I am identifying. think I am actively developing my themes through analyses, so themes can be identified in two or developed into primary ways. We could be taking an inductive approach. So this is a bottom up approach where my analysis is data driven with extract themes grounded within the data. So the codes that I'm developing, the themes that are developing from this codes derive from the content. So I'm not using any pre-existing theory of concepts in the literature to drive my analyses. If I'm interested in people's experience of assessment methods, I'm looking at my data and based on what participants have said through my coding, I identify and I develop my themes, so it's Data driven. I'm not guided by a particular theory. On the other hand, I might be taking a deductive approach towards thematic analysis, which is an example from mine and Laura's paper that I showed. This is where I'm using existing theory to guide my analyses and the development of themes, so I can be analysing my data through the lens of pre-existing theory, pre-existing concepts. I could be looking at my data using my data to explore particular theoretical ideas. So in the paper that I've shown you, looking at people's experiences and representations of the illness, as a health psychologist, I am aware of key models in the literature. So one particular model is the common sense model of illness Misrepresentations, which has subsequently changed names. So I'm aware that that individuals represent and understand their illness along a number of dimensions. Now I'm using this knowledge to explore those particular theoretical ideas in my data. That's not to say I'm going through it as a bit of a checklist dimension. Consequences are there be added? Yep. That's there is how those ideas are manifested in the data. So what consequences do individuals talk about when it comes to Covid 19, for example, and so forth? So I'm using existing theory to guide that analyses. Okay. So. We'll be covering thematic analyses using the approach that was outlined by Braun and Clark, firstly in 2006, and they have published subsequent papers and books following that initial publication. And it provides a nice, organised, structured way to conduct your analyses. And they outline that thematic analysis through their approach involves six particular steps. Now these are sequential. Each builds from the previous one. But that's not to say once you've done with generating initial codes and you've gone on to develop your candidate themes, that's that you don't go back. In fact, you do go back and forth in the process until you develop your final set of themes, and you define and you name them and you write your narrative. So take you through each of those steps with some examples. And we'll also do some activity associated with initial coding. So we've got familiarising yourself with the data, generating some initial code, searching for themes, reviewing themes, defining and naming, and finally producing your report. Now the first step. If you're working with verbal data, if you've contacted your interviews, if you've conducted your focus groups and so forth, the first thing that you would need to do is transcribe that data into written form. To do thematic analysis, we need text. We need words. Now I brought an example. Oh yes I have. So this is an example. Focus group transcript from Bronwen Clark To illustrate what your transcript would likely look like. So you've done, for example, your focus groups. You've audio recorded them with participants consent. I always mention ethics with participants consent. The next step is to get the the audio into written words. You can do this manually. There are software that you can use as well to help, but there's always some manual work that you need to do, for example, checking that everything is accurate, but also you present what was said and who said it, which is quite important. Module is the moderator said this. Sally said this. Rebecca said that. Now one important thing to point out here, these are likely pseudonyms. So not participants real names to maintain the Confidentiality and anonymity of participants. Another way to do it it could be researcher participant one, participant two, and so forth. So this is what your focus group transcript would look like if you contact. Um, an online survey with open text questions you wouldn't need to transcribe that is already in written text. So thinking last week when I spoke about when deciding your methods depends on your research questions and the resources that you've got at the time as well. So the first step is you really need to get to know your data and broaden. And Clark used the word I'm immersing in the data. So reading your data set at least a couple of times from beginning to end. Five focus groups read those focus group transcripts from beginning to end. It's really important at this stage that we really do familiarise ourselves with our data and it's active reading, looking at the data analytically, what are participants telling us, what does that mean? And so forth. We might end this phase by making a few notes on some of the ideas that are becoming evident in our data, and what's important to note here, these are just some initial ideas, some initial notes that you're making. They're not codes. They're not themes. At this stage, we shouldn't be skipping a step. So the next step is for you to generate some initial codes. Now when we talk about codes, codes are typically short phrases that capture the essence of why you think an extract from a focus group, from an interview, from a survey response is interesting and relevant in relation to your research question. Now, you could go about this in a number of ways. I've got my focus group transcript. I read it from beginning to end. I identify a piece of text that's relevant to my question. I could do this on pen and paper. I highlight the text and I write a short label, a short code next to it that captures the essence of what that extract is all about. And I'll talk about semantic and latent coding shortly. It's important that we're quite thorough with our coding code. Everything that you feel is important is relevant to your question at this stage. At a later stage, you can make some decisions as to why do I keep all of the codes? Do I discard some of the codes? You could do this electronically as well. There are software and vivo to help with coding and organising. You could be using Excel or Word and typing your code any way that works for you. I still do most of my things with pen and paper. It suits the way I work and then I put everything on the computer. So it's a process of identifying aspects of the data that relate to your research question. We have to be inclusive. We have to be thorough working through each data set item before proceeding to the next. Now codes provide the building blocks of our analyses. If I imagine my analyses as a house with walls, brick walls, and tiled roofs, those walls and the roof, those are my themes and the individual bricks. The individual tiles are the codes that I use to develop those themes. So your approach to coding depends on the type of thematic analyses. You can be coding at the semantic or at the latent level. So remember earlier when I said we need to make some active choices, some decisions when we conduct thematic analyses. This is one of them. So if I'm developing semantic codes, those are codes that are essentially a relabel that provides a summary of the explicit of the semantic content of the data, what participants have said. So I'm not trying to interpret what they've said. I'm not looking at any underlying meanings. It's a summary of the semantic, the surface level meaning of the data. Latent codes. On the other hand, this is where we go beyond the explicit content, and we look at identifying meanings that lie beneath the words. And this is where, as a researcher, I bring in my conceptual, my theoretical frameworks to try and identify those implicit meanings within the data. So just to illustrate with an example again from Vaughn and Clark, let's try and find an example of a semantic code. If we look at this extract this piece of text from I think it was a focus group. Yes it was. I think modern technology like allows you to be lazy as well because you don't have to do things for yourself. You can get machines and stuff to do things for you. Now, this particular code is an example of a semantic code. So we maps onto the content of what the participant has said. If we take another one for example, kids don't know how to cook. If we look at what Carla has said and then their children are growing up not knowing the faintest idea to even cook or prepare food. Again, this is an example of a semantic code. And when we take this approach, it's important that our codes are in fact semantic in nature. This is an example of a latent code. So humans are naturally lazy. Now this goes if we look at the relevant extract. This goes beyond the explicit content of the data. So participants never actually express this directly and explicitly, but many of the things that they say. Around exercise, around modern lifestyles, rely on this particular understanding of what humans might be like. So it reflects assumptions, frameworks, etc. views underpinning that data. Now, a couple of things that I like to mention here from my experience as well with coding. So. Particular extracts can be coded in as many ways as it's fit for purpose. Relevant to my research question. For example, if I take what Judi is telling me here, Judi pseudonym Judi, this has been coded in three different ways, as you can see, each Capturing different elements in the data that might be meaningful in relation to my research question. It could be the case that some extracts are not coded at all, because they might have no relevance to my question. Or it could be the case that in a large chunk I can only develop one code that's relevant. Question. Yes. UNKNOWN So the thing is saying that our technology facilitates cohesity. SPEAKER 1 But what Judy is saying when she says can do things for you. Um, the thing is, is. SPEAKER 0 That just to make sure that this is up and running as well, the context is absolutely important, but also when thinking about what makes a good code on his own. First, I would say that codes need to be as concise as possible. They need to capture the essence of what it's about, that bit of your data that interests you in relation to your research question. And they have to be informative enough that they stand alone on their own. They should have the right amount of context, because ideally you'd want codes not to be specific to one particular individual situation. You want to be able to use that code again. If that same idea is present in other places within your data set. Okay. So we'll do a 5 to 7 minute activity where hopefully. SPEAKER 2 You'll be able to see this. SPEAKER 0 I will put it on Moodle as well. So let's focus on the third cell here. What Sally is saying where I've put my cursor either individually or in pairs or small groups. However you want to do this, absolutely fine. Hover, read through that data, extract and try and develop 1 or 2 semantic codes. Okay. And then we'll come back and have a bit of a conversation about this. At this stage also we can exclude codes from going further in the analyses if they're not relevant to our research question. So it's really important that at this stage we collate our codes, because that's what we'll be using for the next step in our analyses. So this next step involves organising the different codes into potential themes. So you will start with your long list of collated codes. You review those codes and you try and you consider the different ways in which you can combine them into particular candidate themes at this stage. For example, are there any broad topics, um, concepts, issues that particular codes which similar codes relate to? Can you group your codes together expressing similar ideas? So that's the process where we look at our codes and we try and structure them, group them in meaningful ways into broader themes. Now you start to explore the relationship between themes and how themes work together at this stage. And what you end up with is candidate themes, and this is really important. This is not your final set of themes we still need to review. We still need to evaluate and add this next stage as well. You might go back to your coding. You might go back to your extracts and look at the content within those themes. So. Just want to point out that relationships between themes can be non-hierarchical. For example, you can have themes as a result of your thematic analyses. Or you could have a hierarchical structure. For example, in the paper that I presented earlier by myself and Laura Blackie, we had an overarching theme and themes within that. In this particular example that I'm showing, you have an overarching theme of modern life, and this is mainly used to organise and structure your themes, the analyses you tend not to contain, um, data or codes within overarching themes, and then within those you can have themes and subthemes as well, which can capture and develop multiple specific aspects of themes. You don't have to have subthemes. Those develop from the analyses where meaningful. What you're seeking is a simple structure to represent a capture was in your data in relation to the research question. Now, what's important here is that we don't just develop a candidate themes and leave them jobs done. We need to be reviewing our themes. We need to be looking at our themes, both whether the codes within each theme cohere together meaningfully. Are the codes meaningfully related to each other, but also whether each theme is capturing distinct ideas. And this is really important. What we're looking at here is each theme has very clear boundaries. There isn't any overlap in ideas. If I'm talking, for example, about. Let me think of an example. Decision making with regards to the vaccine, and that appears in one particular theme. I can't really be talking about decision making with regards to the vaccine in a different theme as well. So this is a very simple example. It can be more complicated than that. So each theme is distinct in terms of the ideas or the codes that it represents. And what we want is for each theme to be rich with a number of diverse ideas. By ideas I mean codes. So each code represents a unique idea in your data set in relation to the research question. Now this is where we need to go back to our data extracts and our entire data set, and make a judgement of whether those themes work together to tell us a story about the meanings within the data. And again, at this point you might decide, well, two themes. They might not be rich enough. There might be an overlap in ideas. In conceptual ideas we can combine to create a broader theme or some themes you might decide to to divide. Now, once we developed our themes, the next step, if I go back to the example. The next step is to generate some names for the themes. This will capture the essence of what each theme is about and the ideas encapsulated within that theme. So for example, the first theme that we developed with Laura here in relation to the transition back to normal routines, is the living with the uncertainty. So that reflected how individuals so society is progressing towards those final stage of the pandemic and that there was some lingering uncertainty within individuals. So this was my first theme. Now what's important to point out here, if I've got interview questions, if I've got survey questions, those questions are not my themes. I haven't analysed my data appropriately. If my questions are, then my themes. Your themes are identified across the data set across a number of questions. Okay. And what we would then need to do in a subsequent step, once we have defined our themes, is to write an analytic narrative of that theme, to define, to describe, and the relations between the different ideas, the different codes within the theme. So, for example, you see here where I present my results, these are the themes that we've developed. And I then go on. This is one way of presenting your analysis, the write up of your of your results section. You talk about the themes. Name them, define them, and start talking about presenting an analytic narrative. I'm not just paraphrasing what my participants have said. I'm producing a narrative that illustrates the different facets of each theme, and I'm supporting that with relevant quotations from participants responses. So this appear to be distinct. That's just because of the format of this particular journal. But here you can see that I've embedded those quotations as evidence of the ideas of the points that I'm talking about in my theme. Okay, so producing the report and this is another example here of how they've done it in the Malik and Colson paper as well. So my advice would be to have a look at some examples from the literature, from the online resources by Braun and Clarke. They do have some examples of data guided examples, have a go at some coding and developing themes and so forth. And I'm giving you some examples from the literature here and some further reading to support what we've talked about today.