Hi everyone. I'm going to show you how to analyze qualitative data using ChatGPT. Imagine you have interviewed participants, you are gotten all the information that you need and you want to make sense of the information.
What do you have to do? You could use chart G P T to make sense of your data, and I'm going to show you the process. I'm going to show you how to summarize your qualitative data, how to do the summary based on your research question that you have, how to generate initial codes.
And also lastly, I will be showing you how to develop themes. So these are the four things that you're going to learn about. If you have any question, feel free to put in the comments section.
I'll be happy to address them for you. Summarizing your qualitative data. Sometimes you need to do the summer just to get a big picture of what the data is telling you and ChatGPT can help you to do the summary.
Imagine you have interviewed participants who are Primary healthcare physicians - collecting data from them. They are five, this is one of the data. You collected the demographic information and also questions about the experience of burnout, the causes and also the solutions of burnout.
So I have five of these documents, see participant one to five. I have already analyzed the data using qualitative software NVivo. NVivo is a software that you can use to analyze your data.
So as, you can see here I have my research question here and the codes that I've developed. For the first research question and the second research question, I have all the here. As we are going through the process, I'll be comparing what ChatGPT will be giving me to the data I have analyzed.
And see whether there's a relationship or differences between them. Can ChatGPT give me similar results? That's what I want to find out.
So first of all, what you have to do is to do the summary. We are going to copy all the data. So I think they are about one page for each of the transcripts.
So we're gonna have about five pages. So you're gonna copy all of them and put it in ChatGPT. One thing that you have to notice, that ChatGPT can only take about 3000 words.
If you put more than 3000 words there, the system will tell you that it's too much for it to process. But there's a way out. One thing that you could do is that you can send that information in chunks.
You can send maybe first 3000 words, second can 3000 words. Then after that, you tell ChatGPT to review all the information that you have sent, and then, develop a summary for you. But this one I just want to show you how to summarize our data using ChatGPT.
So let's see. So now that we have our data, we are going to go to ChatGPT and copy and paste that information there and see what we are going to get. Don't forget to subscribe to my YouTube channel.
Trying to make sure that I give you all the information that you need to be a very good qualitative researcher, in terms of doing your research and analyzing your data and presenting your findings. So follow my channel and I'll be happy to also address any questions that you have. I have ChatGPT plus.
So I have the two versions. If you have the free version, you have only GPT 3. 5.
So we are going to use the GPT 3. 5. So let's start.
First, what I would do is to put in the prompt. The PowerPoint has all the prompts there for you to practice. So this is the prompt.
Summarize the transcript in thousand words. You can put in a different way. You can say, summarize the transcript below in thousand words.
So the thousand words could be about two pages or so, right? You can say 500 words, you can say 200 words. But I like thousand words because it gives you a lot of information that you could get a big picture of the data that you have.
So now that I've finished writing this one, I'm going to copy and paste all the information from the transcript. Let me copy here. This is the first one I paste here.
And if you click on enter everything you go. So don't click on enter yet. Copy and paste everything first.
I'm looking for the second transcript. Let me open the second transcript and then copy and paste. Normally, I use this form of summarizing to help me to better understand, what exactly the data is all about.
So let's say if you want to get initial understanding of the data, this is the good practice for you to do. So now that I've copied everything. The next step is to click this one or just press or enter on the keyboard.
So you see here that the system has summarized everything for you, and it's about maybe a thousand words I haven't checked yet. I can copy and put it on a Word document to count the words. But as you can see here, the system have provided you information about the first interviewee.
And the second participant, the third participant, fourth participant, and fifth. It has summarized each of the transcripts for you. I think that's very good information because now you have a brief information about each of the participants.
So I like this one. You get an understanding of each of them before even you go into the data and analyze. You don't have to use chatGPT to fully analyze your data.
You can use it now like what I'm doing, to just summarize and learn about the data before you go through line-by-line and manually develop codes to address your research question that you have. So you can use it and at the same time doing the manual coding or using the software. So now it's good you have brief information about each of the participants and instead of you going through all the transcripts and learn about them.
At the beginning of the stage of your data analysis, having this one will be very helpful for you. This kind of summary we call it general summary because we didn't ask ChatGPT to address a question for you. You ask it to just summarize.
Imagine that you have a research question and you go back to ChatGPT to review the transcripts and address the research question that you have. So that's what we're gonna do for the second step. This is the prompt that we are going to use.
Summarize the transcript based on the research question. What are the causes of burnout among primary healthcare physicians? So let's try and see what we are going to get.
We put the prompt here, and let's see. So we can see here that, ChatGPT has provided a summary addressing the research question that we have. So you can see here that they talk about heavy workload categorized by long hours, numerous clinical and administrative task.
So you can see here that the system has provided some of the information about the causes of burnout. This one is also talks about lack of work-life balance. It is also mentioned a coping mechanism, but that's not our focus.
Our focus is more about the causes of burnout. It said overall, the causes of burnout among primary healthcare physicians identify in the transcript include heavy workload, administrative burnout, lack of work life balance, high expectations. .
. that's very excellent. Right?
So ChatGPT is very good in terms of summarizing your data and addressing your research question. You see that there's some similarities between the summary ChatGPT has given us and also what. I analyzed.
Trouble carrying out duties, having long hours, less time for themselves and family, no work family balance. I think that you see the summary covers almost everything that I coded. These are initial codes that I did right when I was coding the data.
So you see how. You could even use as a way of validating the code that you have created, right? So you can analyze the data and use ChatGPT to get a summary, addressing the research question and then you can compare what you found.
I can see that there's a lot of similarities between what I did and also what ChatGPT gave me. Another thing that we can do is that after you have done the summary, addressing the research question, let's go into developing codes. This is where it is going be very interesting because I'll be comparing the calls that I'm going to get with what ChatGPT is going to give me.
Another prompt that we can use is review the transcript to identify relevant excerpts that address the research question. What are the causes of burnout among primary healthcare physicians? Generate phrases, or codes that best represent the excerpts Identified.
Each code should be between two to five words. You know. This one we call it structured prompting, right?
Structured prompting is where you give, the system adequate information for you to get what you want. You tell the system to look at the transcript, right? So every answer that the system is going to give you should be based on the transcript that you have provided.
Then you tell the system that address this question for me, and you ask the system to develop codes addressing this research question. Also you want the code to be in between two to five words. So you see how you're provided a lot of information for ChatGPT to get the information that you need.
So let's try and see what you're going to get. So we copy this one and paste it into ChatGPT and see what you are going to get. This is interesting.
This is what a ChatGPT has given you. It extracted accepts from the transcript and also developed codes. So you see how interesting it is?
ChatGPT goes through the data extracts significant information that can address the research question and develops codes for you and see all these codes. We have a lot of codes here. We have about 20 codes.
Another thing that we have to really look into is that we have to verify whether the significant information that the system has given us is truly from the transcript. Sometimes ChatGPT can make things up. You look here, ChatGPT may provide inaccurate information about people, places or facts.
So you have to be very careful. Don't over depend on the output. You always have to verify.
So one way of verifying is going back to your transcript and see whether the significant information were truly from the data. It's not like the system made it up. So, let's see whether I can find this one from the data.
So the first participant, oh, it looks like it's here. You see, long hours, numerous clinical and administrative tasks. So this shows that ChatGPT directly extracted that information from the transcript.
Let's try the second one, a lack of balance with work and home life. Oh, I think it's here. You see a lack of balance with work and home life.
So just checking too. I think I now have confidence in the significant information the system has extracted. So, you always have to check because you don't want to just take this information and start presenting it to your audience without checking.
So another thing that we can also check is the code that ChatGPT provided. Let's try to compare it with what I did. Although my codes are a little bit different in terms of the construction, they are quite similar.
Work life balance. I have something like no work family balance. So there are some similarities concerning what I developed and also the system developed.
So I think that it looks like ChatGPT is doing a great job concerning making sense of data, so you see how you can work hand in hand with ChatGPT, especially when you reach a stage where you're having difficulty coming up with a code, right? So you can use ChatGPT first to generate codes, and then you use the codes to go through the data that you have. So it helps you with the beginning to start the coding process.
So that's what I would use if I were to analyze my qualitative data. So now that we have been successful developing codes, let's go ahead and do something interesting. What we could do to follow a systematic process is to ask ask system to review the codes and generate themes out of the codes, right?
Or another option is to. ask system to use the summary to generate themes to help you to address your research question. I'm thinking that if you feel satisfied with the initial code s ChatGPT has given you, then you can go ahead and ask ChatGPT to categorize the codes for you so that it can develop themes for you.
I'm fully satisfied with what ChatGPT has given me. So what we are going to do is that we are going ask ChatGPT to generate themes, addressing the research question based on these codes. So we can say, can you categorize the codes and generate themes addressing the research question.
Let me paste the research question here. Addressing the research question, what are the cause of burnout among primary healthcare physicians? And I said each themes should be in between two to five words.
Okay, so let's see what the system will give us. Okay, so we have seven themes, right? As you can see here, workload and time pressure.
And we have these under workload and time pressure. And then we go to work life imbalance, and we have all these codes. So what you could do is after you have gotten this one, you can review the codes.
Because at the end of the day, you also have to give a human touch to it, right? So if the system gives you this one, don't run with it. Always verify, examine, question the results, and make sure that the codes that are under each of the themes are supposed to be there, right?
So you see how easy the process is. We started with a summary and then we went to initial codes, and now we have asked system to categorize the codes for us to develop themes. And also you can ask the system to put the results in the table form.
So you can say, can you put this recent output in a table form. Let's see what this, see, nicely. The system has put everything in a table form.
You see how amazing ChatGPT is? Putting everything in the table form for you. Let's ask another question.
It's gonna be very interesting. It's getting interesting. Can you add a third column to the table and the column.
The new column should have excerpts from the transcript supporting the themes. Let's see what we're gonna get. Wow.
Wow. You see how, it's interesting that now you have the theme you have codes, you have excerpts from the transcript, but remember, always check whether the excerpts are truly from the transcript. So it's very important for you to check.
You finish your analysis. Amazing work, what you can do with ChatGPT. You can use it, especially when you have a huge amount of data.
You don't have time and resources to analyze all the data or analyze everything. So you can use our gpt as a way of getting your codes, your themes and excerpts, and presenting your findings. Amazing.
I hope this one was so helpful. If you have any question, feel free to put Your question in a comment section, I'll be happy to address them for you. And I hope this one was so helpful.
If you want me to come and speak to your organization about ChatGPT and doing research, I'll be happy to do that. Email me and I'll be happy to respond to your email and if you want a consultation, if you want to learn how to use ChatGPT to analyze your qualitative data, I'm there to help. Enjoy the rest of your day.
Thank you so much for your time.
