187 – An Introduction to Presentation Graphs

[00:00:07] Lucie: Hi, and welcome to the IDEMS podcast. My name’s Lucie Hazelgrove Planel I’m a Social Impact Scientist and anthropologist, and I’m here today with David Stern, one of the founding directors of IDEMS.

Hi, how are you doing?

[00:00:18] David: I am doing well Lucie. Looking forward to a discussion today, what are we talking about?

[00:00:23] Lucie: How to really have impact in a project. So not just thinking about sharing your results from all the sort of basic graphs that you’re doing, but how to really think about presenting your results in a way which is really going to influence people, which is really gonna catch their eye and help them understand what amazing things you have found.

So I think you call them presentation graphs.

[00:00:49] David: Yes. Well, that’s ambitious for what I think of as presentation graphs. I like the ambition, I think of presentation graphs really as a part of three ways you can use descriptive statistics or graphics or data visualisations in your work. One is you have your, I would argue descriptive graphs, your simple descriptives, which can be used for quality control, they’re things you should do automatically when you get data almost.

You then have exploratory graphs, as I would put them, which is when you have a research objective or you have something you are wanting to find out about, and you then dig in and you explore that using data visualisations of various forms.

[00:01:35] Lucie: So this is the distinction, perhaps for example, where if you’ve got a big survey, then your sort of quality control graphs would be perhaps to check that you do have the right number of people.

[00:01:45] David: To understand what missing values there are in your data, to understand what people sort of responded to the different questions, did they respond to the different questions? You know, the sort of standard things. If you have a tool like ODK or Kobo Collect, the things you get for free, this is what you’d actually get often in a dashboard automatically. Those would be what I’d call descriptive graphs. They give you the very quick descriptions of what you’ve collected. And you can automate that and people do.

[00:02:17] Lucie: And some of those descriptive graphs though, you can also classify as exploratory graphs because they do start looking also at, well, depending on if your data is clean, they do start giving you some summaries which are useful.

[00:02:29] David: I’m not saying that descriptive graphs aren’t useful. I’m saying that descriptive graphs you can obtain automatically without knowing what you are trying to study. Whereas your exploratory graphs, this is once you have a particular research objective, you are wanting to bring particular variables together, and you are wanting to explore with respect to gender, with respect to the villages, with respect to other components that you are wanting to do. But that relates to the structure of your study.

So automating that is harder. So your descriptive, your initial descriptives, this is what you can automate easily and just get out. Whatever survey you do, whatever data you collect, you can get your descriptives.

[00:03:17] Lucie: Thank you for discussing that, ’cause I think, yeah, I think, within our team, the RMS team at the moment, I think they’re still progressing from understanding the difference, clearly, I am too, understanding the difference then between those simple descriptions and then the explorations, because the explorations, they’re not yet knowing how to guide their explorations, how to orientate them.

[00:03:40] David: The explorations in my mind, this, the big difference is your descriptives don’t really depend on the meaning of the individual variables and the study behind it, the thinking of the study behind it. Whereas your exploratory descriptives, your exploratory graphs, those are things which you are pursuing, you are exploring within the data based on the objectives you are wanting to study and the structure you’ve got in your data.

[00:04:14] Lucie: Yeah, ’cause in a conversation I had earlier this morning, we weren’t sure, for example, whether the fact of, whether husbands necessarily had the same perspectives as their wives. And so that was something that we were wanting to explore to see, well, is it something that’s interesting or not?

[00:04:29] David: Absolutely, is there anything interesting in the data related to the gender perspective? You know, that’s a really good exploratory sort of question. Now, of course, some of these, such as that gender question, these can be semi-automated where you say, well, of course I actually want to know most of my key variables with respect to are gender differences, and so on.

So there’s lots of powerful things which can be automated or semi-automated. But your exploration in general, this is a process which depends on the objectives of the study and what you’re trying to find out more than just the data, whatever that data is, and be that data quantitative or qualitative.

[00:05:13] Lucie: Okay. And so that’s really helpful. So now we are talking also about these presentation graphs.

[00:05:19] David: For me, the addition you have is not only do you now have the objective and you have maybe results which you are wanting to display, but you also have an audience, you have people to whom you are wanting to communicate those results or the investigation you’ve done.

So that’s the progression that I see. Your descriptives, your initial descriptives, they’re just looking at the data, your exploration adds in objectives, things you’re wanting to find out. And now you found something out, when you now want to communicate that to an audience. And so now you want to present it to that audience.

And that’s the key because, I suppose this is where it comes back to what you were saying at the beginning. It’s not going to be impactful to your audience unless you know what the characteristics are of your audience, which are going to determine what you should have. If you’ve got a very scientific audience, then you might want to have hypothesis tests in there so that you can actually know which of the differences are significant or not significant. If you’ve got a farmer audience, you might want to avoid abstraction, you know, summaries are not as good as actually being able to see individuals.

What’s really interesting is we found working with farmers, directly with farmers, some of whom have been illiterate, they can deal with a lot of complexity, they can deal with complex looking graphs. But they don’t tend to interpret things like averages well, because, well, who’s an average farmer?

Whereas actually seeing, oh yes, there were a number of farmers who got these results and there were other farmers who got these results, my results are here. These are things where even if there’s a lot of complexity, they get a lot out of it. And it’s the same looking at historical rainfall, they can actually interpret a lot of complexity as long as they can find the things they know. Oh yeah, that year was the really wet year, this was the year with lots of droughts and so on.

So, being able to not summarise in ways that particularly lead to abstraction is really important to that sort of audience.

[00:07:30] Lucie: That’s really interesting.

[00:07:31] David: And then of course you’ve got your reporters or your policy makers, they do want you to summarise. They want you to be able to have things which tell…

[00:07:41] Lucie: A really clear message.

[00:07:43] David: Exactly, and making a very clear statement, a very clear message, you are removing ambiguity in terms of this, you’ve got complexity, which is, you know, they’re not wanting to sort of have to grapple with the full complexity, they want to be able to say, well, what’s the key message? And you focus on that key message and making sure that that’s coming out and it’s clearly represented.

[00:08:04] Lucie: Because presentation graphs can also be for academic purposes, I assume.

[00:08:08] David: Absolutely.

[00:08:09] Lucie: Where that complexity, you do often want it, you do want to say there it is a bit complicated.

[00:08:15] David: Well, absolutely. This is what I was saying. If your audiences are scientists, you want not only that complexity but the scientific rigour. You want the hypothesis test to be represented in some way so that that can be actually, that interpretation can be done quickly and efficiently, and people can know, yes, okay, this difference, this is actually a statistically significant difference based on the data you’ve got.

Different audiences will need different presentation graphs of the same results potentially, or the same core results. So thinking about your audience alongside your results, that would be what I consider a presentation graph.

And very often the way I’d sort of see it is your descriptives, these can be automated or they can be done very quickly. Your explorations, you are spending your time looking across many different graphs in quick iteration, as you dig into the exploration you are doing. Your presentation graph, you can be spending hours or even days on a single visualisation, getting it just right for your audience because it’s worth that investment of time to communicate the message you want succinctly.

And often, if you think of the grammar of graphics language, you might want multiple layers to this, where you’ve actually got different bits of data being represented on the same graph to really bring different messages out.

[00:09:46] Lucie: And if it’s for journalists or policy, then it may not even resemble a standard graph, potentially.

[00:09:53] David: Yes, that’s right, you might be bringing maps in, you might be having images, you might be having elements of qualitative and quantitative together on the same visualisation to really bring out the story you are trying to tell in a way which is both, depending on your audience, rigorous enough, but also powerful in its presentation.

[00:10:16] Lucie: So could you give an example of that, of how, you know, you might combine different data sources or something?

[00:10:22] David: Well, I mean, think of the really nice visualisations you’ve seen in different publications. I’ll give an example, which is not a visualiszation I’ve created, but one I could imagine. There’s an island in the Caribbean where their historical rainfall data has an outlier. Every time they try and clean the data, this outlier stands out. And if you didn’t have institutional memory, they would probably throw it away because you never have a rainfall value that high.

But this was a very specific hurricane which happened to hit the mountain behind it at a particular angle, which meant that they just had this real downpour, incredible flooding, more than double the next highest rainfall event. And so if you imagine now having this on a graph, which shows the historical rainfall events across the years, and this one, it comes out and it is sort of standing out like a sore thumb.

[00:11:21] Lucie: Exactly, anyone seeing that would ask themselves questions.

[00:11:23] David: Well, this is where you might then have a quote from, you know, a newspaper article or something saying biggest rainfall event in history, or whatever it is, which then is sort of there and actually pointing towards, I don’t know what the name of the hurricane was, where people who know the history of the island will then say, yes, that was that event.

You could even have a little description about it because that’s something worth highlighting. And it’s worth bringing together this sort of quantitative data, this historical data about rainfall in different ways, with this qualitative data about that individual event which makes the quantitative data unusual, it’s an extreme event in that.

That would be an example of a visualisation I’d love to see. You could include a photo of the devastation from that event because it sort of really is important to draw people’s attention to what is possible or what happened on that one occasion if that’s what you are trying to communicate.

[00:12:26] Lucie: I’m thinking that NGOs often, you know, have a map of the world to show where they work, or international NGOs, they will often also have, you know, photos to sort of show different types of work they do in different types of countries or different countries.

[00:12:38] David: With maybe a bit of a quote about what they’re doing and with maybe number of people they’re reaching or something like this. So this combination of qualitative and quantitative.

[00:12:48] Lucie: Can, we go into that more? Like, I have a big question, the basis of this is really quantitative ’cause it’s, you know, it’s a graph that’s sort of comparing numbers or something. And that can be based on qualitative research because you can quantify different things. I’m wondering, is it always based on a graph?

[00:13:09] David: As a visualisation, you mean?

[00:13:11] Lucie: Yeah.

[00:13:12] David: Think of it in a different way. And there’s some wonderful tools for this where you actually start with a story and you tell your story and then within your story you include graphs and figures and numbers and so on. You’ve got to think about, even if you think about your webpage, you might have that graph on a webpage, well, is the graph the whole or is the story within which the graph sits the whole? And what’s the actual, what’s the message you’re trying to get across? In journalism the same thing holds, your graph fits into a narrative.

The question really that you’re saying, you know, does the visualisation, does it start with the quantitative or the qualitative? Well, it depends how big your visualisation is. What are you actually considering within it? There can be a narrative component, which comes from the story you’re trying to tell, which could come from qualitative research you’ve done, or you could start with, well, there’s actually this quantitative result we’re trying to display and then you are wanting to actually use the qualitative to supplement it. So either of those are possible.

[00:14:17] Lucie: Yeah, that’s interesting, ’cause the story is, the story is a type of visualisation, when it’s written.

[00:14:23] David: And it doesn’t have to be written, these can be very visual. This is the thing, you can tell stories very visually, with maps, with timelines, with you know, other, well, other visual cues, if you want, putting them together. Once you start talking about this, you’re now putting days or weeks into creating a visualisation. This is something which is a substantial piece of work, it’s an art form done well.

And that’s something to be valued in its own right. Yes, you can actually, more and more, create something quite quickly using AI and using the data that you have in different ways. But powerful stories warrant time and skill. That skill set to be able to tell those powerful stories is one which is really important. And it’s not necessarily just the quantitative, there’s plenty of people who have good quantitative results but haven’t been able to tell their story as powerfully. The skill of telling a powerful story is a whole different skillset, which many scientists could benefit from training in this, and actual thinking about this more.

[00:15:42] Lucie: But as you say, part of the key is also having a good story to tell. So if you’ve done research and actually you find that, you know, something’s too variable or something, and so you don’t really understand the situation, then the story isn’t so easy to tell because you need to do more research perhaps. This only really works if someone has done research and they have some really exciting result they want to share.

[00:16:06] David: Well, even that, I would argue it’s, I know some really exciting results which are difficult stories to tell because they are about complexity, and so it comes back to what audience are you wanting to tell this. Your best scientific results, maybe you can tell that story really well to a scientific audience, but not to a popular audience. And there’s other stories which you might be able to tell really well to a popular audience, which scientifically it’s, well, okay, yeah, we kind of knew that, that’s not very interesting.

[00:16:40] Lucie: Yes, that’s a very good point. Yeah.

[00:16:41] David: So, again, it depends on the audience. Now there are sometimes these ability to tell both at once, and when that comes together, it’s beautiful. And what I do find interesting is that that coming together is very often about the skills of the individual.

[00:17:00] Lucie: By telling both at once, sorry, do you mean telling the same story but to different people or using the same…?

[00:17:06] David: Having a good scientific story, which is also a good popular story.

[00:17:12] Lucie: Yeah.

[00:17:13] David: Quite often when this has come together, it is about the individual working on it and the individuals actually being able to see both and work towards both together. Actually having the scientific rigour to get the good scientific result at the same time as having your eye on that good popular story and getting the results for that and drawing those together into a coherent whole, I don’t know that many cases where that’s come together without somebody who is extremely good at that, where somebody’s, where you separate out those two tasks.

[00:17:48] Lucie: Exactly, often people are very much directed at just the research audience or just about working with communities.

[00:17:54] David: Yes, exactly. And it’s those rare individuals who are good at both, I tended to find they’ve been the ones who have been able to bring out some of these. Which I would argue is a reason for looking at building communication skills in good scientists and scientific skills in good communicators.

[00:18:14] Lucie: Yeah. Yeah. Great. Thank you very much for that introduction, David.

[00:18:18] David: You’re welcome. My hope is this is something we are going to be digging into more as part of the community of practice over the next year or so.

[00:18:26] Lucie: Yeah, I’m afraid that’s why I called it an introduction.

[00:18:29] David: Exactly.

[00:18:31] Lucie: Okay. Thanks then

[00:18:33] David: Thanks.