Data Visualization

class: center, middle, inverse, title-slide

.title[
# Data Visualization
]
.subtitle[
## Chapter 1. Fundamentals of Graphical Practice
]
.author[
### Iñaki Úcar
]
.institute[
### Department of Statistics | uc3m-Santander Big Data Institute
]
.institute[
### Master in Computational Social Science
]
.date[
### <p>Licensed under Creative Commons Attribution <a href="https://creativecommons.org/licenses/by/4.0/">CC BY 4.0</a> Last generated: 2023-09-27</p>
]

---

class: base24, middle, clear

- [Introduction to DataViz](ch1_1.html#3)
- [Graphical Integrity](ch1_2.html#3)
- [Graphical Perception](ch1_3.html#3)
- [Principles of Graphical Representation](ch1_4.html#3)

---
class: inverse, center, middle

# Graphical Integrity
## How to Lie with Graphs

---
class: base24

# Bad Charts

![:vspace 50]()

- We live surrounded by bad charts

- For many people, statistical charts = **lie**

![:vspace 30]()

- Deception vs. ineptitude
  - Lack of **quantitative skills**, _anumerism_
  - Lack of knowledge about **visual perception**
  - Lack of **visual skills**
  - Intention to **deceive**

---
class: base24

# Sources of Bad Charts

- For Tufte:

- Distortion
  - Data variation
  - Visual area and numerical measure
  - Context

- But there are many more **common patterns**:

- Bad representation (deception, sometimes)
  - Bad data
  - Too many data (optimism)
  - Too little data (optimism)
  - Hiding relevant data (deception)
  - Wrong representation (deception)
  - Creativity (without skill)

---

# Prologue: Pie Charts

.panelset[
.panel[.panel-name[See the pattern?]
.center[![](assets/img/ch1/s2/bad-pie1.png)]
]
.panel[.panel-name[Second try]
.center[![](assets/img/ch1/s2/bad-pie2.png)]
]
.panel[.panel-name[The only acceptable one]
.center[![](assets/img/ch1/s2/bad-pie3.png)]
]
]

---

# Bad Representation: Areas

.panelset[
.panel[.panel-name[See the pattern?]
.center[![](assets/img/ch1/s2/bad-area1.png)]
]
.panel[.panel-name[Second try]
.center[![](assets/img/ch1/s2/bad-area2.png)]
]
]

---

# Bad Representation: Longitudes

.panelset[
.panel[.panel-name[Deception 1]
- Cutting the origin is the most common deception technique with bar plots

![:vspace 50]()
.pull-left[.center[![](assets/img/ch1/s2/bad-bar1.png)]]
.pull-right[.center[![](assets/img/ch1/s2/bad-bar2.png)]]
]
.panel[.panel-name[Deception 2]
.center[![](assets/img/ch1/s2/bad-bar3.png)]
]
.panel[.panel-name[Ineptitude 1]
.center[![](assets/img/ch1/s2/bad-bar4.png)]
]
.panel[.panel-name[Ineptitude 2]
- Some well-intentioned axis cuts may have nefarious consequences

![:vspace 50]()
.pull-left[.center[![](assets/img/ch1/s2/bad-bar5.png)]]
.pull-right[.center[![](assets/img/ch1/s2/bad-bar6.png)]]
]
]

---

# Bad Data

.panelset[
.panel[.panel-name[Population maps]
.center[![](assets/img/ch1/s2/bad-data1.png)]
]
.panel[.panel-name[Irrelevant data]
.center[![](assets/img/ch1/s2/bad-data2.jpg)]
]
.panel[.panel-name[Bonus]
.center[![:scale 80%](assets/img/ch1/s2/bad-data3.jpg)]
]
.panel[.panel-name[Non-comparable]
.center[![](assets/img/ch1/s2/bad-data4.png)]
]
.panel[.panel-name[Not related]
.center[![](assets/img/ch1/s2/bad-data5.png)]

- See [Spurious correlations](https://www.tylervigen.com/spurious-correlations) for more fun
]
]

---

# Too Many Data

.panelset[
.panel[.panel-name[Pie horror 1]
.center[![](assets/img/ch1/s2/bad-many1.jpg)]
]
.panel[.panel-name[Pie horror 2]
.center[![](assets/img/ch1/s2/bad-many2.png)]
]
]

---

# Too Little Data

.panelset[
.panel[.panel-name[Optimist 1]
.center[![](assets/img/ch1/s2/bad-little1.png)]
]
.panel[.panel-name[Optimist 2]
.center[![](assets/img/ch1/s2/bad-little2.png)]
]
.panel[.panel-name[Scientific optimist]
.center[![](assets/img/ch1/s2/bad-little3.png)]
]
.panel[.panel-name[+bad interpretation]
.center[![](assets/img/ch1/s2/bad-little4.png)]
]
]

---

# Hiding Relevant Data

.panelset[
.panel[.panel-name[No context]
![:vspace 50]()
.center[![](assets/img/ch1/s2/bad-hide1.png)]
]
.panel[.panel-name[More context]
![:vspace 50]()
.center[![](assets/img/ch1/s2/bad-hide2.png)]
]
.panel[.panel-name[Complete picture]
![:vspace 50]()
.center[![](assets/img/ch1/s2/bad-hide3.png)]
]
.panel[.panel-name[Context!]
.center[![](assets/img/ch1/s2/bad-hide4.png)]
]
]

---

# Wrong Representation

.panelset[
.panel[.panel-name[Level 1]
.center[![](assets/img/ch1/s2/bad-wrong1.png)]
]
.panel[.panel-name[Level 2]
.center[![](assets/img/ch1/s2/bad-wrong2.png)]
]
.panel[.panel-name[Level 3]
.center[![](assets/img/ch1/s2/bad-wrong3.png)]
]
.panel[.panel-name[Level 4]
.center[![](assets/img/ch1/s2/bad-wrong4.jpg)]
]
.panel[.panel-name[Level 5]
.center[![](assets/img/ch1/s2/bad-wrong5.jpg)]
]
.panel[.panel-name[Level 6]
.center[![](assets/img/ch1/s2/bad-wrong6.jpg)]
]
]

---

# Creativity

.panelset[
.panel[.panel-name[Level 1]
.center[![](assets/img/ch1/s2/bad-creativity1.png)]
]
.panel[.panel-name[Level 2]
.center[![](assets/img/ch1/s2/bad-creativity2.png)]
]
.panel[.panel-name[Level 3]
.center[![](assets/img/ch1/s2/bad-creativity3.png)]
]
.panel[.panel-name[Level 4]
.center[![](assets/img/ch1/s2/bad-creativity4.png)]
]
.panel[.panel-name[Level 5]
.center[![](assets/img/ch1/s2/bad-creativity5.png)]
]
.panel[.panel-name[Level 6]
.center[![:scale 60%](assets/img/ch1/s2/bad-creativity6.png)]
]
]

---

# Bonus: Scientific Horrors

.panelset[
.panel[.panel-name[3D]
.center[![](assets/img/ch1/s2/bad-scientific1.jpg)]
]
.panel[.panel-name[Linear relationship?]
.center[![](assets/img/ch1/s2/bad-scientific2.jpg)]
]
.panel[.panel-name[Quadratic relationship?]
.center[![:scale 60%](assets/img/ch1/s2/bad-scientific3.jpg)]
]
.panel[.panel-name[Other?]
.center[![:scale 60%](assets/img/ch1/s2/bad-scientific4.jpg)]
]
.panel[.panel-name[Whatever]
.center[![:scale 60%](assets/img/ch1/s2/bad-scientific5.jpg)]
]
]

---

# 12 Rules to Display Data Badly

.footnote[Wainer, H. (1984) **How to Display Data Badly**. _The American Statistician_ 38(2), pp. 137-147.]

.pull-left[
### Showing data

- Rule 1. Show as few data as possible (minimize the data density)
- Rule 2. Hide what data you do show (minimize the data-ink ratio)
]

.pull-right[
### Showing data accurately

- Rule 3. Ignore the visual metaphor altogether
- Rule 4. Only order matters
- Rule 5. Graph data out of context
]

### Showing data clearly

- Rule 6. Change scales in mid-axis
- Rule 7. Emphasize the trivial (ignore the important)
- Rule 8. Jiggle the baseline
- Rule 9. Austria first!
- Rule 10. Label illegibly, incompletely, incorrectly, and ambiguously
- Rule 11. More is murkier: more decimal places and more dimensions
- Rule 12. If it has been done well in the past, think of another way to do it

---

# Chart Rethinking: Exercise 1

.footnote[https://junkcharts.typepad.com]

.panelset[
.panel[.panel-name[Original]
.center[![:scale 60%](assets/img/ch1/s2/rethinking11.jpg)]
]
.panel[.panel-name[Iteration 1]
.center[![:scale 60%](assets/img/ch1/s2/rethinking12.png)]
]
.panel[.panel-name[Iteration 2]
.center[![:scale 52%](assets/img/ch1/s2/rethinking13.png)]
]
]

---

# Chart Rethinking: Exercise 2

.footnote[https://junkcharts.typepad.com]

.panelset[
.panel[.panel-name[Original]
.center[![:scale 70%](assets/img/ch1/s2/rethinking21.png)]
]
.panel[.panel-name[Iteration 1]
.center[![](assets/img/ch1/s2/rethinking22.png)]
]
.panel[.panel-name[Iteration 2]
.center[![:scale 45%](assets/img/ch1/s2/rethinking23.png)]
]
]

---

# Chart Rethinking: Exercise 3

.footnote[https://www.datanalytics.com/]

.panelset[
.panel[.panel-name[Original]
.center[![](assets/img/ch1/s2/rethinking31.png)]
]
.panel[.panel-name[Original]
.center[![:scale 40%](assets/img/ch1/s2/rethinking32.png)]
]
.panel[.panel-name[Iteration 1]
.center[![:scale 40%](assets/img/ch1/s2/rethinking33.png)]
]
]

---

# Chart Rethinking: Exercise 4

.footnote[https://junkcharts.typepad.com]

.panelset[
.panel[.panel-name[Original]
.center[![:scale 33%](assets/img/ch1/s2/rethinking41.png)]
]
.panel[.panel-name[Iteration 1]
.center[![:scale 62%](assets/img/ch1/s2/rethinking42.png)]
]
]

---

# Chart Rethinking: Exercise 5

.footnote[https://junkcharts.typepad.com]

.panelset[
.panel[.panel-name[Original]
.center[![](assets/img/ch1/s2/rethinking51.png)]
]
.panel[.panel-name[Iteration 1]
.center[![:scale 75%](assets/img/ch1/s2/rethinking52.png)]
]
.panel[.panel-name[A word of caution]
.center[![](assets/img/ch1/s2/rethinking53.png)]
]
]

---

# Chart Rethinking: Exercise 6

.footnote[Schwabish, J. A. (2014). _An Economist’s Guide to Visualizing Data_]

.panelset[
.panel[.panel-name[Original]
.center[![:scale 78%](assets/img/ch1/s2/rethinking61.png)]
]
.panel[.panel-name[Iteration 1]
.center[![:scale 75%](assets/img/ch1/s2/rethinking62.png)]
]
]

---

# Chart Rethinking: Exercise 7

.footnote[Schwabish, J. A. (2014). _An Economist’s Guide to Visualizing Data_]

.panelset[
.panel[.panel-name[Original]
.center[![:scale 78%](assets/img/ch1/s2/rethinking71.png)]
]
.panel[.panel-name[Iteration 1]
.center[![:scale 43%](assets/img/ch1/s2/rethinking72.png)]
]
]

---

# Chart Rethinking: Exercise 8

.footnote[Schwabish, J. A. (2014). _An Economist’s Guide to Visualizing Data_]

.panelset[
.panel[.panel-name[Original]
.center[![:scale 97%](assets/img/ch1/s2/rethinking81.png)]
]
.panel[.panel-name[Iteration 1]
.center[![:scale 84%](assets/img/ch1/s2/rethinking82.png)]
]
]

---
class: base24

# Summary

- We live surrounded by **bad charts**, either by deception or ineptitude.

- There are some **common patterns** to be aware of.<br>
  From those, we can distill a series of **basic rules**:

- Pie charts are very rarely justified.
  - Areas are tricky.
  - 3D is rarely justified in 2D mediums.
  - Axes (cuts, scales...) should be clear.
  - Proper context should be shown when relevant.
  - Do not go beyond what data tells.
  - There are **well-established visual metaphors**.
      - E.g. up-down for positive-negative.
      - E.g. areas are perceived as whole quantities,<br>
        so cutting the axis is deceiving.