Even beautiful maps can be misleading

# Even beautiful maps can be misleading
## How decisions about spatial data visualisation affect map legibility
### Sam Langton & Réka Solymosi
### NSCR & University of Manchester
### 19/01/2022

---

???

Image credit: [Andisheh Nouraee](https://twitter.com/andishehnouraee/status/1284237474831761408)

---
class: center, middle

# Maps: visual representation of spatial data

---

# About us

Réka Solymosi

- [@r_solymosi](https://twitter.com/r_solymosi)

- [rekadata.net](https://rekadata.net/)

![UoM Logo](http://assets.manchester.ac.uk/logos/lo-res/UniLlogoTable_whitebackgrounds.jpg)

![SSI Logo](https://www.software.ac.uk/themes/ssi/images/ssi_logo_with_name-small2.png)

]

--
.pull-right[

Sam Langton

- [@sh_langton](https://twitter.com/sh_langton)

- [samlangton.info](https://www.samlangton.info/)

![NSCR Logo](https://nscr.nl/app/themes/nscr/out/assets/images/logos/logo_NSCR-EN.svg)

]

---
class: inverse, center, middle

# Why maps?

---
class: center, middle
# Much of human activity happens 'somewhere'

---
# So we see maps everywhere

![](img/bbcmap.png)

???

Image credit: [BBC News](https://www.bbc.com/news/world-51235105)

]

---

# Some are very useful!

Optimal Manchester brewery crawl (see [https://www.ncalvert.uk/posts/drunkensalesman/](https://www.ncalvert.uk/posts/drunkensalesman/))

![](https://www.ncalvert.uk/images/OptimalBreweryCrawl.png)

---

# But maps can pose issues

![](https://imgs.xkcd.com/comics/heatmap.png)
]

???

Image credit: [XKCD](https://xkcd.com/1138/)

---

# Sometimes bad maps are funny

---
# Some people collect bad maps

---
class: center, middle

# Key considerations

--
### - What is the intended message?
--

### - How do different maps convey this message?

---

# Two key fields

--
.pull-left[

### - cartography

![](https://upload.wikimedia.org/wikipedia/commons/thumb/d/d6/Kepler-world.jpg/250px-Kepler-world.jpg)

]

### - (non-spatial) data visualisation

![](https://upload.wikimedia.org/wikipedia/commons/thumb/1/17/Nightingale-mortality.jpg/120px-Nightingale-mortality.jpg)
]

???

Image credit: [Wikimedia commons](https://commons.wikimedia.org/wiki/Category:Coxcomb_(Florence_Nightingale))

---
class: inverse, center, middle

# Empirical understanding of how people perceive different "viz"

---
# Example 1: Pie charts

--
![](https://eagereyes.org/wp-content/uploads/2016/07/simplevis-abserror-relative.png)

Image credit: [Robert Kosara](https://eagereyes.org/blog/2016/a-reanalysis-of-a-study-about-square-pie-charts-from-2009)

---
# Example 2: Tufte charts

--
.center[
![](img/chartjunk.png)
]

Image credit: [Bateman, S., Mandryk, R. L., Gutwin, C., Genest, A., McDine, D., & Brooks, C. (2010, April). Useful junk? The effects of visual embellishment on comprehension and memorability of charts. In Proceedings of the SIGCHI conference on human factors in computing systems (pp. 2573-2582).](https://dl.acm.org/doi/abs/10.1145/1753326.1753716)

---
# Examples in cartography

--
.center[
![](img/gradsymb1.png)

]

Image credit: [Flannery, JJ (1971) The relative effectiveness of some common graduated point symbols in the presentation of quantitative data](https://files.eric.ed.gov/fulltext/ED045469.pdf)

---
class: inverse, middle, center

# Addressing a specific problem: variation in the size and shape of areas

---

# Eg: USA States

# Eg: UK Local Authorities

# Message often obscured

.center[
![](mod_presentation_files/figure-html/unnamed-chunk-4-1.png)
]
---
# A fix: distort polygons

---

---
class: inverse, center, middle

# How does this affect perception?

---
# Our focus: EU referendum results

---

# 4 types of distortions

---

# (a) Balanced cartogram

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

- Scale by min value to balance **invisibility** and **distortion**

- Eg if SIU is 0.02 inches on a 5inch map:

```r
siu <- 0.02 # the smallest interpretable unit
height <- 5
bb <- sp::bbox(map)
width <- (bb[1,2] - bb[1,1]) / (bb[2,2] - bb[2,1]) * height
bbA <- (bb[1,2] - bb[1,1]) * (bb[2,2] - bb[2, 1])
mapA <- rgeos::gArea(map)
minA <- (siu * bbA) / (height * width)
map$scaleby <- rgeos::gArea(map, byid = TRUE)
map$scaleby[map$scaleby < minA] <- minA
# Use this to scale cartogram
balcarto <- cartogram::cartogram(map, "scaleby", maxSizeError = 1.1, prepare = "none")
```

---

![](img/map_a.png)
]
---

# (b) Hexogram

- same idea as balanced carto, with minimum value being what allows each area to be represented as its own hexagon

- hexagons are produced using the hexagonal binning function in R’s fMultivar package, based on the centroids of each polygon

- our example:

```r
# Get the function needed
script <- RCurl::getURL("https://raw.githubusercontent.com/profrichharris/Rhexogram/master/functions.R")
eval(parse(text = script))
# Number of bins guided by the -binN- function for a visual plot.
# 29 is also used by Harris in example.
harris.29  <- hexogram(LAE.sp, 29)
# Extract the hexograms
harris.29.hex.sp <- harris.29[[2]]   # 2 is hexo
```

---

![](img/map_b.png)

]

---

# (c) geogrid squre grid

- From [Joseph Bailey](https://cran.r-project.org/web/packages/geogrid/readme/README.html)

- Calculates a grid that strives to preserve the original geography.

- 2 steps to using this

- 1 - Generate grid by varying the seed
  
  - 2 - Efficiently calculate the assignments from the original geography to the new geography
  
--

- Our example:

```r
# step 1 generate grid
LAE.reg <- calculate_grid(shape = LAE.sp, grid_type = "regular", seed = 1) #1 was our fave seed
# step 2 calculate assignments
LAE.reg <- assign_polygons(LAE.sp, LAE.reg) 
```

---

![](img/map_c.png)
]

---

# (d) geogrid hexagonal grid

- Same idea as with grids but for hexagons:

```r
# step 1 generate grid
LAE.hex <- calculate_grid(shape = LAE.sp, grid_type = "hexagonal", seed = 1)

# step 2 calculate assignments
LAE.hex <- assign_polygons(LAE.sp, LAE.hex)
```

---

![](img/map_d.png)
]

---
.center[
# "High values (in yellow) appear to be clustered near one another, with a handful of outliers elsewhere in the country"
]

### - 5-point Likert scale (strongly agree, slightly agree, neither agree nor disagree, slightly disagree, strongly disagree).

- Higher agreement = better representation of statement in map

- Convenience sample (internet) of 768 respondents

---

# Results

]
---

# Results (contd.)

![](mod_presentation_files/figure-html/unnamed-chunk-10-1.png)

]

---

# Conclusions

- New methods to visualise geographic information **can** convey a message more accurately than original thematic maps.

- But choose the method with consideration to the research question and the data!

---

# Crime Mapping

Forthcoming textbook based on our UoM Crime Mapping module:

---

# Thanks!

- The paper: [Langton, S. H., & Solymosi, R. (2019). Cartograms, hexograms and regular grids: Minimising misrepresentation in spatial data visualisations. Environment and Planning B: Urban Analytics and City Science.](https://journals.sagepub.com/doi/full/10.1177/2399808319873923)

- These slides: [https://rekadata.net/talks/mod_presentation.html](https://rekadata.net/talks/mod_presentation.html)

- Questions? Contact Réka ([@r_solymosi](https://twitter.com/r_solymosi)) or Sam ([@sh_langton](https://twitter.com/sh_langton))

.small[
- Slides created via the R package [**xaringan**](https://github.com/yihui/xaringan). The chakra comes from [remark.js](https://remarkjs.com), [**knitr**](http://yihui.org/knitr), and [R Markdown](https://rmarkdown.rstudio.com).
]