how humans see data

205
How Humans See Data John Rauser @jrauser November 2016

Upload: john-rauser

Post on 14-Apr-2017

1.280 views

Category:

Technology


0 download

TRANSCRIPT

Page 1: How Humans See Data

How Humans See Data

John Rauser@jrauser

November 2016

Page 2: How Humans See Data

How Humans See Data

John Rauser@jrauser

November 2016

Page 3: How Humans See Data

visualization

Page 4: How Humans See Data

visualizationis

communication

Page 5: How Humans See Data

how to make better visualizations

Page 6: How Humans See Data

help humans solve analytical problems quickly and accurately

with visualization

Page 7: How Humans See Data

Part I: Why visualize data at all?

Page 8: How Humans See Data
Page 9: How Humans See Data

x1.972

y1.236

x y

0.111 0.5421.112 1.994 0.902 0.0050.000 1.009 0.598 0.0850.665 1.942 1.613 1.7900.235 0.356 1.298 1.9550.247 1.658 0.651 1.9371.275 1.961 1.949 1.3160.702 0.045 0.099 0.5671.760 0.350 0.862 0.0101.691 0.277 0.027 0.7681.628 1.778 0.706 1.9561.957 1.290 1.042 1.999

Page 10: How Humans See Data
Page 11: How Humans See Data

pre-attentive processing

Page 12: How Humans See Data

A graph is an encoding of the data.

Page 13: How Humans See Data

x1.972

y1.236

x y

0.111 0.5421.112 1.994 0.902 0.0050.000 1.009 0.598 0.0850.665 1.942 1.613 1.7900.235 0.356 1.298 1.9550.247 1.658 0.651 1.9371.275 1.961 1.949 1.3160.702 0.045 0.099 0.5671.760 0.350 0.862 0.0101.691 0.277 0.027 0.7681.628 1.778 0.706 1.9561.957 1.290 1.042 1.999

Page 14: How Humans See Data
Page 15: How Humans See Data

n x y n x y1 1.972 1.236 13 0.111 0.5422 1.112 1.994 14 0.902 0.0053 0.000 1.009 15 0.598 0.0854 0.665 1.942 16 1.613 1.7905 0.235 0.356 17 1.298 1.9556 0.247 1.658 18 0.651 1.9377 1.275 1.961 19 1.949 1.3168 0.702 0.045 20 0.099 0.5679 1.760 0.350 21 0.862 0.010

10 1.691 0.277 22 0.027 0.76811 1.628 1.778 23 0.706 1.95612 1.957 1.290 24 1.042 1.999

Page 16: How Humans See Data
Page 17: How Humans See Data
Page 18: How Humans See Data

Good visualizations optimize for the human visual system.

Page 19: How Humans See Data

How does the human visual system work?

Page 20: How Humans See Data

How does the human visual system decode a graph?

Page 21: How Humans See Data
Page 22: How Humans See Data

Cleveland’s three visual operations of pattern perception:

1. Detection2. Assembly3. Estimation

Page 23: How Humans See Data

Part II: estimation

Page 24: How Humans See Data

Three levels of estimation

a. discrimination X=Y X!=Yb. ranking X>Y X<Yc. ratioing X / Y = ?

Page 25: How Humans See Data

At the heart of quantitative reasoning is a single question: Compared to what?

- Tufte, Envisioning Information

Page 26: How Humans See Data

Three levels of estimation

a. discrimination X=Y X!=Yb. ranking X>Y X<Yc. ratioing X / Y = ?

Page 27: How Humans See Data
Page 28: How Humans See Data
Page 29: How Humans See Data

the most important

thing

Page 30: How Humans See Data
Page 31: How Humans See Data

The most important measurement should exploit the highest ranked encoding possible.

• Position along a common scale• Position on identical but nonaligned scales• Length• Angle or Slope• Area• Volume or Density or Color saturation• Color hue

Page 32: How Humans See Data

The most important measurement should exploit the highest ranked encoding possible.

• Position along a common scale• Position on identical but nonaligned scales• Length• Angle or Slope• Area• Volume or Density or Color saturation• Color hue

Page 33: How Humans See Data

The most important measurement should exploit the highest ranked encoding possible.

• Position along a common scale• Position on identical but nonaligned scales• Length• Angle or Slope• Area• Volume or Density or Color saturation• Color hue

Page 34: How Humans See Data

“The first rule of color: do not talk about color!”

- Tamara Munzner

Page 35: How Humans See Data

luminance

saturation

hue

Page 36: How Humans See Data

luminance

saturation

hue

Page 37: How Humans See Data
Page 38: How Humans See Data
Page 39: How Humans See Data
Page 40: How Humans See Data
Page 41: How Humans See Data

Observation: Alphabetical is almost never the correct ordering

of a categorical variable.

Page 42: How Humans See Data
Page 43: How Humans See Data
Page 44: How Humans See Data

The most important measurement should exploit the highest ranked encoding possible.

• Position along a common scale• Position on identical but nonaligned scales• Length• Angle or Slope• Area• Volume or Density or Color saturation• Color hue

Page 45: How Humans See Data
Page 46: How Humans See Data
Page 47: How Humans See Data
Page 48: How Humans See Data
Page 49: How Humans See Data
Page 50: How Humans See Data

The most important measurement should exploit the highest ranked encoding possible.

• Position along a common scale• Position on identical but nonaligned scales• Length• Angle or Slope• Area• Volume or Density or Color saturation• Color hue

Page 51: How Humans See Data
Page 52: How Humans See Data
Page 53: How Humans See Data
Page 54: How Humans See Data

The most important measurement should exploit the highest ranked encoding possible.

• Position along a common scale• Position on identical but nonaligned scales• Length• Angle or Slope• Area• Volume or Density or Color saturation• Color hue

Page 55: How Humans See Data
Page 56: How Humans See Data
Page 57: How Humans See Data
Page 58: How Humans See Data
Page 59: How Humans See Data

The most important measurement should exploit the highest ranked encoding possible.

• Position along a common scale• Position on identical but nonaligned scales• Length• Angle or Slope• Area• Volume or Density or Color saturation• Color hue

Page 60: How Humans See Data

11 mpg

Page 61: How Humans See Data

11 mpg

Page 62: How Humans See Data

11 mpg

Page 63: How Humans See Data
Page 64: How Humans See Data

The most important measurement should exploit the highest ranked encoding possible.

• Position along a common scale• Position on identical but nonaligned

scales• Length• Angle or Slope• Area• Volume or Density or Color saturation• Color hue

Page 65: How Humans See Data
Page 66: How Humans See Data
Page 67: How Humans See Data
Page 68: How Humans See Data

The most important measurement should exploit the highest ranked encoding possible.

• Position along a common scale• Position on identical but nonaligned scales• Length• Angle or Slope• Area• Volume or Density or Color saturation• Color hue

Page 69: How Humans See Data
Page 70: How Humans See Data
Page 71: How Humans See Data

The most important measurement should exploit the highest ranked encoding possible.

• Position along a common scale• Position on identical but nonaligned scales• Length• Angle or Slope• Area• Volume or Density or Color saturation• Color hue

Page 72: How Humans See Data

Observation: Stacked anything is nearly always

a mistake.

Page 73: How Humans See Data
Page 74: How Humans See Data
Page 75: How Humans See Data
Page 76: How Humans See Data
Page 77: How Humans See Data
Page 78: How Humans See Data

Stacking makes the reader decode lengths, not position

on a common scale.

Page 79: How Humans See Data

11 mpg

Page 80: How Humans See Data
Page 81: How Humans See Data

Observation: Stacked anything is nearly always

a mistake.

Page 82: How Humans See Data
Page 83: How Humans See Data

Observation: Pie charts are

ALWAYS a mistake.

Page 84: How Humans See Data

Piecharts are the information visualization equivalent of a roofing hammer to the frontal lobe. They have no place in the world of grownups, and occupy the same semiotic space as short pants, a runny nose, and chocolate smeared on one’s face. They are as professional as a pair of assless chaps.

http://blog.codahale.com/2006/04/29/google-analytics-the-goggles-they-do-nothing/

Page 85: How Humans See Data

Piecharts are the information visualization equivalent of a roofing hammer to the frontal lobe. They have no place in the world of grownups, and occupy the same semiotic space as short pants, a runny nose, and chocolate smeared on one’s face. They are as professional as a pair of assless chaps.

http://blog.codahale.com/2006/04/29/google-analytics-the-goggles-they-do-nothing/

Page 86: How Humans See Data

The most important measurement should exploit the highest ranked encoding possible.

• Position along a common scale• Position on identical but nonaligned scales• Length• Angle or Slope• Area• Volume or Density or Color saturation• Color hue

Page 87: How Humans See Data
Page 88: How Humans See Data
Page 89: How Humans See Data

Tables are preferable to graphics for many small data sets. A table is nearly always better than a dumb pie chart; the only thing worse than a pie chart is several of them, for then the viewer is asked to compared quantities located in spatial disarray both within and between pies… Given their low data-density and failure to order numbers along a visual dimension, pie charts should never be used.

-Edward Tufte, The Visual Display of Quantitative Information

Page 90: How Humans See Data

Tables are preferable to graphics for many small data sets. A table is nearly always better than a dumb pie chart; the only thing worse than a pie chart is several of them, for then the viewer is asked to compared quantities located in spatial disarray both within and between pies… Given their low data-density and failure to order numbers along a visual dimension, pie charts should never be used.

-Edward Tufte, The Visual Display of Quantitative Information

Page 91: How Humans See Data

Clinton TrumpAmong Democrats 99% 1%Among Republicans 53% 47%

Who do you think did a better job in tonight’s debate?

Page 92: How Humans See Data
Page 93: How Humans See Data

Afghanistan Albania Algeria Angola ArgentinaAustralia Austria Bahrain Bangladesh BelgiumBenin Bolivia Bosnia and Herzegovina Botswana BrazilBulgaria Burkina Faso Burundi Cambodia CameroonCanada Central African Republic Chad Chile ChinaColombia Comoros Congo, Dem. Rep. Congo, Rep. Costa RicaCote d'Ivoire Croatia Cuba Czech Republic DenmarkDjibouti Dominican Republic Ecuador Egypt El SalvadorEquatorial Guinea Eritrea Ethiopia Finland FranceGabon Gambia Germany Ghana GreeceGuatemala Guinea Guinea-Bissau Haiti HondurasHong Kong, China Hungary Iceland India IndonesiaIran Iraq Ireland Israel ItalyJamaica Japan Jordan Kenya Korea, Dem. Rep.Korea, Rep. Kuwait Lebanon Lesotho LiberiaLibya Madagascar Malawi Malaysia MaliMauritania Mauritius Mexico Mongolia MontenegroMorocco Mozambique Myanmar Namibia NepalNetherlands New Zealand Nicaragua Niger NigeriaNorway Oman Pakistan Panama ParaguayPeru Philippines Poland Portugal Puerto RicoReunion Romania Rwanda Sao Tome and Principe Saudi ArabiaSenegal Serbia Sierra Leone Singapore Slovak RepublicSlovenia Somalia South Africa Spain Sri LankaSudan Swaziland Sweden Switzerland SyriaTaiwan Tanzania Thailand Togo Trinidad and TobagoTunisia Turkey Uganda United Kingdom United StatesUruguay Venezuela Vietnam West Bank and Gaza Yemen, Rep.Zambia Zimbabwe

Page 94: How Humans See Data

All good pie charts are jokes.

Page 95: How Humans See Data
Page 96: How Humans See Data

Observation: Comparison is trivial on a common scale.

Page 97: How Humans See Data
Page 98: How Humans See Data
Page 99: How Humans See Data
Page 100: How Humans See Data
Page 101: How Humans See Data
Page 102: How Humans See Data

the dashboard metaphor is fundamentally flawed

Page 103: How Humans See Data
Page 104: How Humans See Data
Page 105: How Humans See Data
Page 106: How Humans See Data

Observation: Scatterplotsshow relationships directly.

Page 107: How Humans See Data
Page 108: How Humans See Data
Page 109: How Humans See Data

Observation: Growth charts usually aren’t.

Page 110: How Humans See Data
Page 111: How Humans See Data

If growth (slope) is important, plot it directly.

Page 112: How Humans See Data
Page 113: How Humans See Data

Observation: Growth charts usually aren’t.

If growth (slope) is important, plot it directly.

Page 114: How Humans See Data

The most important measurement should exploit the highest ranked encoding possible.

• Position along a common scale• Position on identical but nonaligned scales• Length• Angle or Slope• Area• Volume or Density or Color saturation• Color hue

Page 115: How Humans See Data

Cleveland’s three visual operations of pattern perception:

1. Detection2. Assembly3. Estimation

Page 116: How Humans See Data

Part three: assembly

Page 117: How Humans See Data

Gestalt Psychology

Page 118: How Humans See Data
Page 119: How Humans See Data

reification

Page 120: How Humans See Data

emergence

Page 121: How Humans See Data
Page 122: How Humans See Data

emergence

Page 123: How Humans See Data
Page 124: How Humans See Data

Prägnanz

Page 125: How Humans See Data

Law Of Closure

Page 126: How Humans See Data
Page 127: How Humans See Data

Law Of Continuity

Page 128: How Humans See Data
Page 129: How Humans See Data
Page 130: How Humans See Data

Observation: Good plots leverage the law of continuity

to assist with assembly.

Page 131: How Humans See Data
Page 132: How Humans See Data
Page 133: How Humans See Data

Law of Similarity

Page 134: How Humans See Data
Page 135: How Humans See Data
Page 136: How Humans See Data
Page 137: How Humans See Data
Page 138: How Humans See Data
Page 139: How Humans See Data
Page 140: How Humans See Data

Law of Proximity

Page 141: How Humans See Data
Page 142: How Humans See Data
Page 143: How Humans See Data

Observation: dodged bar charts are a bad idea

Page 144: How Humans See Data

Cleveland’s three visual operations of pattern perception:

1. Detection2. Assembly3. Estimation

Page 145: How Humans See Data

Part IV: detection

Page 146: How Humans See Data
Page 147: How Humans See Data
Page 148: How Humans See Data
Page 149: How Humans See Data
Page 150: How Humans See Data
Page 151: How Humans See Data
Page 152: How Humans See Data
Page 153: How Humans See Data

excel’s defaults are pretty bad

Page 154: How Humans See Data

1 2 3 4 5 6 -

20,000

40,000

60,000

80,000

100,000

120,000

140,000

160,000

180,000

200,000

Page 155: How Humans See Data

Observation: Detection isn’t as trivial as it seems.

Page 156: How Humans See Data
Page 157: How Humans See Data

“Above all else, show the data.”-Tufte

Page 158: How Humans See Data

Part V: other useful results

Page 159: How Humans See Data

Weber’s law: The “Just Noticeable Difference” is proportional to the

size of the initial stimuli.

Page 160: How Humans See Data

10 20

Page 161: How Humans See Data

10 20

100 110

Page 162: How Humans See Data
Page 163: How Humans See Data
Page 164: How Humans See Data

12 units

12 units

Page 165: How Humans See Data

Observation: Weber’s Law is why gridlines are useful

Page 166: How Humans See Data
Page 167: How Humans See Data
Page 168: How Humans See Data
Page 169: How Humans See Data

“Erase non-data ink.”

-Tufte

Page 170: How Humans See Data

“Erase non-data ink, within reason.”

-Tufte

Page 171: How Humans See Data

“Erase non-data ink that interferes with detection or doesn’t assist assembly and estimation.”

-Rauser

Page 172: How Humans See Data

You are best at detecting variation in slope near 45 degrees.

Page 173: How Humans See Data
Page 174: How Humans See Data

banking to 45

Page 175: How Humans See Data
Page 176: How Humans See Data
Page 177: How Humans See Data

Observation: Banking to 45 best shows variation in slope

Page 178: How Humans See Data
Page 179: How Humans See Data

Q: Should I include 0 on my scale?

Page 180: How Humans See Data
Page 181: How Humans See Data
Page 182: How Humans See Data

Q: Should I include 0 on my scale?

A: It depends.

Page 183: How Humans See Data

Q: Should I include 0 on my scale?

A: Relying on the pre-attentive perception of size or intensity?Yes, otherwise you will mislead.

Using position? It’s up to you.

Page 184: How Humans See Data
Page 185: How Humans See Data
Page 186: How Humans See Data
Page 187: How Humans See Data
Page 188: How Humans See Data
Page 189: How Humans See Data
Page 190: How Humans See Data

“Above all else, show the data.”

-Tufte

Page 191: How Humans See Data

“Above all else, show the variation in the data.”

-Rauser (via Tufte)

Page 192: How Humans See Data

R/GGplot2 code for every plot in this presentation available at http://goo.gl/xH5PLV

The rendered document is at http://rpubs.com/jrauser/hhsd_notes

This presentation is at http://goo.gl/VKxxya

I will tweet these links as @jrauser

Page 193: How Humans See Data

coda

Page 194: How Humans See Data

visualization is

communication

Page 195: How Humans See Data

art is

communication

Page 196: How Humans See Data

visualization is art

Page 197: How Humans See Data
Page 198: How Humans See Data
Page 199: How Humans See Data
Page 200: How Humans See Data
Page 201: How Humans See Data
Page 202: How Humans See Data

why does it make you feel that way?

Page 203: How Humans See Data

visualization has as much to learn from art as from science

Page 204: How Humans See Data

R/GGplot2 code for every plot in this presentation available at http://goo.gl/xH5PLV

The rendered document is at http://rpubs.com/jrauser/hhsd_notes

This presentation is at http://goo.gl/VKxxya

I will tweet these links as @jrauser

Page 205: How Humans See Data

end