Skip to main content
SAGE
Search form
PDF
  • 00:00

    [MUSIC PLAYING]

  • 00:14

    REKA SOLYMOSI: Hi, my name is Dr. Reka Solymosi.And I'm a Lecturer of Quantitative Methodsat the University of Manchester.I'm also a Fellow of the Software SustainabilityInstitute.And before starting my role in academia,I used to be an analyst working with Transport for London.A lot of my research focuses on behavior phenomenon,

  • 00:37

    REKA SOLYMOSI [continued]: elements that take place in a particular placein a particular time.And so spatial visualization of these dataplays a large role in my everyday work.One thing that I learned when working in this areais that not only is it important to analyze the data properly,but to present-- to communicate these findings in ways

  • 00:58

    REKA SOLYMOSI [continued]: that are meaningful.And that the decisions-- the datavisualization decisions that we makeaffect people's interpretation of these results.So more recently, I've worked on a piece of researchwhere we looked at quantifying exactly how the different datavisualization choices might affect

  • 01:19

    REKA SOLYMOSI [continued]: the conclusions that people come away from your analysis.In this video, I'll be talking about that research.So in this particular case study,we were interested in how the decisions that we

  • 01:40

    REKA SOLYMOSI [continued]: make about data visualizations of the waysthat we're presenting the findingscan affect people's interpretations of our message.And to do this, we took one data set,which was a data set of results of the 2016 EU referendum.And we visualized the percentage of people

  • 02:03

    REKA SOLYMOSI [continued]: who voted to remain in each local authorityarea in England.And we found that when visualizing thiswith a choropleth map, you saw small areas disappearand large areas--large as a function of their geography,

  • 02:25

    REKA SOLYMOSI [continued]: really dominate these maps.And that's why we decided to explore this particular casestudy and look at different approachesthat researchers have taken in the past for sortof augmenting these maps.So different visualization, spatial visualizationapproaches that people have taken.And what difference that means for the conclusions

  • 02:47

    REKA SOLYMOSI [continued]: that people might draw about where people voted to remainas part of the EU or leave.So one of the most common ways to visualizespatial data across an area, like across local authorities,

  • 03:09

    REKA SOLYMOSI [continued]: for example, is to use a choropleth map.So this is a type of thematic mapwhere each region is shaded accordingto its particular value on a variable.So you might have something like, in this case,the percentage of people who voted to remain in the EU

  • 03:30

    REKA SOLYMOSI [continued]: and the different local authorities of England.And so this is a continuous scaleand you have this continuous color scheme to represent that,but you can also have a choropleth mapwhere you have discrete categories and discrete values.And so your color scheme for shading these mapswill reflect the different discrete categories

  • 03:53

    REKA SOLYMOSI [continued]: that this categorical variable might take.In the case of a choropleth map, each area-- the shapeand size of each area is actuallyreflective of the geographical featuresof this particular region.So in the case of local authorities, whatwe're representing is the geographical shape and sizeof each local authority.

  • 04:15

    REKA SOLYMOSI [continued]: And this might have some benefits,and that we're staying true to the original geography.But it also has some drawbacks.And as you can see in the case of visualizingthe percentage of people who voted to remain in the EuropeanUnion, we had a cluster of peoplewho really wanted to remain who concentrated in London.

  • 04:38

    REKA SOLYMOSI [continued]: But the local authority areas in London, they'requite small geographically.Especially when we're zooming outto visualize all of England.And so this concentration is really difficultto notice and to see in this particular choiceof visualization.And you might think that this isn't a decision that we makebecause we're not changing their geographies,

  • 04:59

    REKA SOLYMOSI [continued]: we're not making any changes to the shapes,but actually not doing something is still a decisionthat you take when you're visualizing your data.So taking the default. In this case,I guess the default of the geographical shapeand size of the regions is still a decisionthat you're making in your visualization.

  • 05:25

    REKA SOLYMOSI [continued]: One approach to address this problem wheresmall areas can disappear on a mapand large areas really dominate is to use a cartogram.So a cartogram is a type of augmentationto the spatial areas, the polygons on your map, whereyou distort their size and shape in order

  • 05:47

    REKA SOLYMOSI [continued]: to achieve some outcome.For example, you might have a cartogramwhere you weight the size of a particular regionby its population, for example.So you might have seen these really distorted mapsof the world where you see Canada shrinkingto be really, really small.

  • 06:08

    REKA SOLYMOSI [continued]: Because instead of showing the-- representingthe geographic shape of Canada, itis now transformed to represent also this other variableof population.In these distorted map cases, there'san attempt to maintain the neighboringmyths of these polygons, so the contiguity.

  • 06:30

    REKA SOLYMOSI [continued]: But there are other types of cartograms as well.For example, the Dorling Cartogram in which caseeach polygon is replaced with a circledrawn around the centroid of that polygon.And then the size of that circle is augmented,depending on the particular variable.So let's say population.

  • 06:51

    REKA SOLYMOSI [continued]: So a recent development in cartogram visualizationsis the balanced area cardiogram.In the balanced area cardiogram, insteadof taking a variable such as the population size,we use something called the smallest interpretable unitto weight the size and the shape of these areas.

  • 07:12

    REKA SOLYMOSI [continued]: So that is basically how small you can possiblymake one of your polygons, one of your shapesand get away with it while that stillallows the map to be legible.One example given is that if you have a map that's5 inches tall, then maybe 0.2 inchesis the smallest possible interpretable unit.

  • 07:34

    REKA SOLYMOSI [continued]: And it's using these metrics instead of somethinglike population size that we now try and balancethe size of these units of analysisin order to create this balanced area cartogram.In the case study of visualizing the EU referendum results,we used a balance cartogram to augment the size and shape

  • 07:58

    REKA SOLYMOSI [continued]: of the local authorities.And we achieved a map where you can nowsee London a lot better, and you canstart to get an idea of this concentration of the higherpercentage of people who voted to remain sort of clusteringaround a London area.And some of the other more dominating

  • 08:18

    REKA SOLYMOSI [continued]: larger polygons that had high-- lower percentage to remain,so higher percentage to leave, they shrink a little bit.So they no longer visually dominate your map that much.

  • 08:38

    REKA SOLYMOSI [continued]: So when using cartograms to augmentthe shape of the underlying polygons,so in this case, the local authorities,we introduce some distortion, which may stillbe distracting to our viewers.One way of addressing this is to use hexagrams.So hexagrams take the idea behind the balanced cartogram

  • 08:60

    REKA SOLYMOSI [continued]: of the smallest interpretable unit.And with that, we create bins in whichto put the centroids of the polygons of our shapesfrom our balanced hexagram.And it is using the centroid we then apply hexagons,instead of the shape of the cartogram

  • 09:22

    REKA SOLYMOSI [continued]: of the augmented polygons we achieved with the cartogram.And we represent each local authority area with a hexagon.And the interesting thing with this approachis that there is an attempt by the hexagramto try and maintain some of the elements

  • 09:42

    REKA SOLYMOSI [continued]: of the underlying geography.So we want to be as truthful as possibleto the actual locations of these local authoritiesrather than concentrate on tessellation.So because of this, you might see the introductionof some white spaces.So you might see some holes appear in England,

  • 10:05

    REKA SOLYMOSI [continued]: but that doesn't mean there's nothing there.It's just-- or that nobody there votedeither to leave or to remain.It's just that by nature of how we are positioningthese hexagons, they try and stayclose to where the underlying geography wasin the context of this spatial map of England.

  • 10:29

    REKA SOLYMOSI [continued]: So in this case, each local authorityis represented by a hexagon.And you remove some of the distortionthat you get with the balanced cartogram.And you definitely don't have this differencein size and shape that you had with the choropleth map.

  • 10:53

    REKA SOLYMOSI [continued]: The final approach we tested in orderto understand how the different visualizations can affectpeople's interpretations of our messageis to look at grid maps.So grid maps, they attempt to addressthis problem of the varying sizes and shapes

  • 11:14

    REKA SOLYMOSI [continued]: of our polygons, in this case, the local authorities,by substituting these shapes with somethingthat we can overlay as a grid.So for example, squares or hexagons.And by transforming these shapes to squares or hexagons, again,we address the variation in size and shape

  • 11:35

    REKA SOLYMOSI [continued]: that you get with the choropleth map.In this particular case, how thisis different to a hexagram, for example?What we do is we use distance based algorithmsin order to figure out where to put each new shape.So where to move the different local authorities of London

  • 11:59

    REKA SOLYMOSI [continued]: in order that they become nicely tessellatingsquares or hexagons.So in this case, you will see they neighbor one another.But then, what we're doing is tryingto minimize the distance moved.But in the cases of areas where youhave lots of local authorities kind of clustered together,

  • 12:19

    REKA SOLYMOSI [continued]: this might introduce some distortion.So you might see in the cases of these grid maps,that the different local authorities of London, whichyou recognize from their shading from the high percentageof people who have voted to remain in the European Union,have shifted quite far off from where

  • 12:40

    REKA SOLYMOSI [continued]: we would expect them to be with our geographic knowledgeof where is London.So what we wanted to do here is apply two approaches.One grid and one hexagonal approaches to these grid maps,and also compare how people perceive the referendum resultswith these two approaches.

  • 13:09

    REKA SOLYMOSI [continued]: So after producing these five different visualizations-- sowe had the choropleth map of local authoritiesin England, and the percentage thatvoted to remain in the 2016 EU referendum.And then we also had the balanced area cartogram,the hexagram, and then a square grid map,

  • 13:30

    REKA SOLYMOSI [continued]: and hexagonal grid map.And we produced these visualizationswith the same exact underlying data.So all of the maps, they should be telling the same storyif the decisions that we make in visualizing themare not going to affect what the data tells us.

  • 13:50

    REKA SOLYMOSI [continued]: So let's say, we wanted to communicatethat people in London voted to remain in the European Union.So the cluster of votes--the cluster of areas with the higher percentage votingto remain is in a specific part of the country,

  • 14:14

    REKA SOLYMOSI [continued]: is in a specific region.Then does it matter which approach to visualizationswe use?To test this, we took our maps and weran a online crowdsourced survey,receiving over 600 responses from peoplewho viewed these maps and were asked to what extent they agree

  • 14:38

    REKA SOLYMOSI [continued]: with the following statement.High values, in yellow, appear to be clusterednear one another with a couple of outliersscattered elsewhere.So we thought that this statement representsthe data, which shows that the high values, in this case,meaning a higher percentage of people voting to remain,

  • 14:59

    REKA SOLYMOSI [continued]: they clustered in one specific area.So this was the message we were trying to communicatewith the different maps.So we asked people to what extentthey agree with this message having viewed these five maps.So how much they agree with the messagewith each one of these maps.The idea was that if there's no difference,then people should have no difference

  • 15:19

    REKA SOLYMOSI [continued]: in the extent to which they agree or disagree.Because the underlying data doesn't change.But if visualization does make a difference,then we can demonstrate that using a different approachmight affect to what extent people drawthe conclusions that we would want them to from our data.So looking at our results, we cansee that there is actually a difference

  • 15:42

    REKA SOLYMOSI [continued]: in the extent to which people agreed with the statement,depending on what map they were looking at.So with the original map where we didn't do anything,the choropleth map, we didn't change the sides and the shapeof local authorities.We still have people agreeing--mostly agreeing with our statement.

  • 16:02

    REKA SOLYMOSI [continued]: So that's pretty good.This is the statement we want to communicateand people tend to agree with it.We did have some disagreements.So we did have people answering strongly disagree and disagree,but the majority of our respondentsagreed with the message we were trying to communicate.Now compared to the original map,both the balanced cartogram and the hexagram

  • 16:24

    REKA SOLYMOSI [continued]: did significantly better.So a lot more people strongly agreedand agreed with our statement than with the original map.And much fewer people strongly disagreed or disagreedwith our statement.So it does seem that using the balanced areacartogram or the hexagram meant that people

  • 16:46

    REKA SOLYMOSI [continued]: were more likely to agree with the messagethat we were trying to communicate.On the other hand, looking at the grid maps, both the squareand the hexagon grid map seem to perform much worsethan our original local authority map.So then if we hadn't introduced any sort of augmentation

  • 17:07

    REKA SOLYMOSI [continued]: at all.In those cases both, we had more people strongly disagreeand more people disagree than we didwith the original choropleth map.To further quantify this result, wedivided people who said that they agree or stronglyagree into one group of people who agree with our statement,

  • 17:29

    REKA SOLYMOSI [continued]: and people who said they disagree or strongly disagreeinto another group.So that they disagree with their statement.And we ran logistic regression in orderto understand whether for each map type,people are more likely to agree with their messagethan to disagree.Looking at the results from our regressionshows that people were much more likely to agree

  • 17:52

    REKA SOLYMOSI [continued]: with our statements when they saw the balanced areacartogram or the hexagram than when they sawthe original choropleth map.On the other hand, they were much more likely to disagree.So much less likely to agree with the statementthan seeing the original map if they were shown the grid map--

  • 18:17

    REKA SOLYMOSI [continued]: either the square grid map or the hexagonal grid map.So this shows further evidence that the type of visualizationthat we choose to represent our datahas a significant effect on whether peopledraw the conclusions that we intend them to,or whether they might be misled by the choice of visualization

  • 18:40

    REKA SOLYMOSI [continued]: rather than anything to do with our data--underlying data or analysis itself.So what this study showed us is that the choiceof visualization has an effect on the conclusions

  • 19:02

    REKA SOLYMOSI [continued]: that your audiences make about your message.So we looked at the types of decisionsthat you might make in order to addressdifferent issues around the size and the shape of your polygonsin your specific map.So we started with a choropleth mapwhere we visualize the actual shape

  • 19:25

    REKA SOLYMOSI [continued]: and size of the local authority in this case,but whatever would be your spatial unit of analysis.And then we introduce shading to communicate what isthe variation in our variable.But then we saw there are some issues with thatof small areas disappearing, and welooked at four different ways to address that issuewith cartograms.

  • 19:46

    REKA SOLYMOSI [continued]: And specifically, we looked at balanced cartograms,with hexagrams, and then also using gridswith squares and with hexagons.We saw that some of these methodsperformed better than others in getting peopleto agree with the message that we were trying to communicate.And we saw that therefore, the decisionsthat we make in visualization affect

  • 20:09

    REKA SOLYMOSI [continued]: the extent to which people agree with our conclusions.Does this mean that you should justbe using balanced area cartogramsand hexagrams from now on?Nope, that's definitely not the takeaway messagefrom this video.Instead what to focus on is how youthink that the decisions that you're making

  • 20:29

    REKA SOLYMOSI [continued]: and the changes that you're introducing to your mapor not introducing might affect the conclusionsfor your specific data set.If you want to follow along and recreate any of these maps,they were all made using our code.And the code will be available through an appendix for you

  • 20:50

    REKA SOLYMOSI [continued]: to download and follow along.Thank you very much for watching this video,and I hope you enjoy making maps.

Video Info

Publisher: SAGE Publications, Ltd.

Publication Year: 2021

Video Type:Case Study

Methods: Data visualization, Geospatial data

Keywords: accuracy in communication; cartograms; choropleth maps; data visualisation; geospatial data; graphical presentation of data; maps and map-making; research findings ... Show More

Segment Info

Segment Num.: 1

Persons Discussed:

Events Discussed:

Keywords:

Abstract

Dr. Reka Solymosi, PhD, a lecturer of quantitative methods at the University of Manchester and fellow of the Software Sustainability Institute, discusses four data visualization techniques—choropleth maps, cartograms, hexagrams, and grid maps—and their use in minimizing misrepresentation of data.

Video Info

Title:
Minimizing Misrepresentation in Spatial Data Visualizations
Methods:
Data visualization, Geospatial data
Duration:
00:21:23
Discipline:
Video Type:
Case Study
Keywords:
accuracy in communication; cartograms; choropleth maps; data visualisation; geospatial data; graphical presentation of data; maps and map-making; research findings ... Show More

Publication Info

Publisher:
SAGE Publications, Ltd.
Publication Year:
2021
Product:
SAGE Research Methods: Data Visualization
Publication Place:
United Kingdom
SAGE Original Production Type:
SAGE Case Studies
ISBN:
9781529774047
DOI
https://dx.doi.org/10.4135/9781529774047
Copyright Statement:
(c) SAGE Publications Ltd., 2021

People

Academic:
Reka Solymosi

Segment Info

Title:

Segment Num: 1

Keywords:

Segment Start Time:

Segment End Time:

People

Things Discussed

Organizations Discussed:

Events Discussed:

Places Discussed:

Persons Discussed:

Resources

    References
    References

    Further Reading

    Smith, MJ, Goodchild, MF, & Longley, PA(2009).Geospatial Analysis: A Comprehensive Guide to Principles, Techniques and Software Tools. Matador,

    Further Reading

    Harris, R., (2017).Hexograms: Better maps of area-based data. https://rpubs.com/profrichharris/hexograms

    Journal Article

    Brewer, C., (2006).Basic mapping principles for visualising cancer data using geographic information systems. American Journal of Preventative Medicine, 30(2s),s25-s36.

    Journal Article

    Langton, SH, & Solymosi, R(2019).Cartograms, hexagrams and regular grids: Minimising misrepresentation in spatial data visualisations. Environment and Planning B, 48(2),348-357. http://dx.doi.org/https://doi.org/10.1177/2399808319873923

    Conference Proceedings

    Harris, R, Charlton, M, Brunsdon, C, & et al, (2017).Tackling the curse of cartograms. Proceedings of the 25th GIS research UK conference, http://huckg.is/gisruk2017/GISRUK_2017_paper_29.pdf

    Journal Article

    Harris, R, Charlton, M, Brunsdon, C, & et al, (2017).Balancing visibility and distortion: Remapping the results of the 2015 UK general election. Environment and Planning A, 49(9),1945-1947. http://dx.doi.org/https://doi.org/10.1177/0308518X17708439

Methods Map

Data visualization

Presenting data in a visual form (such as a graph, picture or graphics) to showcase and communicate the findings of the research.
Data visualization
Minimizing Misrepresentation in Spatial Data Visualizations

Dr. Reka Solymosi, PhD, a lecturer of quantitative methods at the University of Manchester and fellow of the Software Sustainability Institute, discusses four data visualization techniques—choropleth maps, cartograms, hexagrams, and grid maps—and their use in minimizing misrepresentation of data.

Copy and paste the following HTML into your website