Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Data Visualization for Academic Writing

9332028514858a918dc4af231d4cb4bc?s=47 James Davenport
February 21, 2018

Data Visualization for Academic Writing

A talk given to the Scientific Writing Workshop graduate class (Astr597), taught by Prof. Dalcanton at the University of Washington

9332028514858a918dc4af231d4cb4bc?s=128

James Davenport

February 21, 2018
Tweet

Transcript

  1. Data Visualization for Academic Writing James R. A. Davenport NSF

    Astronomy & Astrophysics Postdoctoral Fellow, Western Washington University DIRAC Fellow, University of Washington jradavenport 1
  2. about me jradavenport

  3. 3 jradavenport

  4. Aurora Borealis - Frederic Edwin Church (1865) jradavenport

  5. jradavenport

  6. jradavenport

  7. jradavenport

  8. jradavenport

  9. jradavenport

  10. jradavenport

  11. Data Visualization or Visual Storytelling jradavenport

  12. “The best statistical graph ever drawn…” -Edward Tufte (except it

    wasn’t drawn for PowerPoint or computers) jradavenport
  13. “The%thing%I've%spent%more%2me%on…% is%trying%to%iden2fy%all%those%things%in% peoples%minds%that%serve%as%obstacles% to%them%understanding”% =%Al%Gore% jradavenport

  14. jradavenport

  15. CO 2$ jradavenport

  16. CO 2$ Temperature$ jradavenport

  17. omg$ CO 2$ Temperature$ omg jradavenport

  18. omg$ CO 2$ Temperature$ A great visualization: sells the message,

    worth 1000 words (or $1.2 Trillion annually) omg jradavenport
  19. What I won’t cover: - how to make graphs -

    types of graphs/plots - graphing languages What I want to cover: - why we make plots - why we publish plots - 5 points to consider jradavenport
  20. Why do we make plots? a.k.a. charts, graphs, visualizations, figures,

    diagrams, maps… jradavenport
  21. Why do we make plots? • Development • data exploration

    • idea generation • debug code • Presentation • talks • posters • papers Each use/format has unique concerns! jradavenport
  22. Anscombe’s Quartet All panels have the same - mean (in

    X and Y) - variance (in X and Y) - linear regression - “r” coefficient Development: Exploratory Data Analysis made popular by Tufte jradavenport
  23. Presentation: Talks & Posters jradavenport

  24. A Tale of One Plot jradavenport

  25. Exploration jradavenport

  26. Finished Product jradavenport

  27. Match Kepler to Gaia (DR1/TGAS) Select Main Sequence Filter out

    “junk” Davenport (2017) jradavenport Lecture Slide
  28. Poster jradavenport

  29. ApJ Webpage jradavenport

  30. ApJ Printed jradavenport

  31. ApJ Printed B/W jradavenport

  32. Your plots (need to) tell the story at a glance!

    Point #1 paper poster jradavenport
  33. One plot does not work for all audiences/needs jradavenport

  34. The Dimensions of Art 65,000 pieces of art from the

    Tate Modern Width Height Width jradavenport
  35. The Dimensions of Art 65,000 pieces of art from the

    Tate Modern jradavenport
  36. One plot does not work for all audiences/needs jradavenport

  37. One plot does not work for all audiences/needs 1st LIGO

    detection, Abbott et al. (2016) LIGO MAGAZINE First detection! LIGO Hanford signal 9:50:45 UTC, 14 September 2015 LIGO Livingston signal issue 8 3/2016 LIGO Scientific Collaboration LIGO Magazine, 2016 jradavenport
  38. One plot does not work for all audiences/needs Davenport et

    al. (2016) blog post jradavenport
  39. Make multiple versions of a plot! Point #2 & make

    them available for people to use! jradavenport
  40. Your camera doesn’t matter – Ken Rockwell (Photographer) A brief

    aside…. jradavenport
  41. A brief aside…. Your graphing language doesn’t matter – me,

    just now A short example: jradavenport
  42. A brief aside…. Can you tell which is IDL, Python,

    Excel ? Kepler data from Davenport et al. (2014) jradavenport
  43. Don’t tool shame Point #3 jradavenport

  44. DESIGN DESIGN DESIGN DESIGN DESIGN jradavenport

  45. A handy design acronym: PARC Proximity Alignment Repetition Contrast Gwen

    Eadie, DIRAC
  46. A handy design acronym: PARC Proximity Alignment Repetition Contrast https://mdst485class.wordpress.com/2016/06/16/this-is-c-r-a-p/

    Gwen Eadie, DIRAC
  47. Keep it Clean !0.025& !0.02& !0.015& !0.01& !0.005& 0& 0.005&

    0.01& 0.015& 0.02& 120& 120.5& 121& 121.5& 122& 122.5& 123& 123.5& Series1& Series1' (0.025' (0.02' (0.015' (0.01' (0.005' 0' 0.005' 0.01' 0.015' 120.5377519' 120.7625304' 120.8238336' 120.885137' 120.9464403' 121.0077433' 121.0690467' 121.1303499' 121.1916532' 121.2938252' 121.3551285' 121.4164317' 121.477735' 121.5594728' 121.620776' 121.6820793' 121.7433826' 121.8046857' 121.8864234' 121.9477266' 122.0090298' 122.0703332' 122.1316365' 122.1929397' 122.254243' 122.3155463' 122.3768494' 122.4381527' 122.499456' 122.5607592' 122.6220625' 122.6833658' 122.744669' 122.8059722' 122.8672756' 122.9490132' 123.0103164' 123.0716198' Series1' Default plots from Excel jradavenport
  48. Keep it Clean Clean plots from Excel and Matplotlib (Python)

    jradavenport
  49. Avoid Gimmicks jradavenport

  50. Avoid Gimmicks examples given for matplotlib… jradavenport

  51. 0 50 100 150 200 April May June July Make

    things comparable jradavenport
  52. Be Consistent jradavenport

  53. Color in visualizations: best friend, worst enemy jradavenport

  54. jradavenport

  55. jradavenport

  56. jradavenport

  57. jradavenport

  58. jradavenport

  59. jradavenport

  60. jradavenport

  61. jradavenport

  62. jradavenport

  63. Try to use good design principles, especially contrast. Point #4

    jradavenport
  64. Attention to Details jradavenport

  65. “… when data understanding is supported by perceptual rather than

    cognitive processes” — Enrico Bertini Visualization works … Daniela Huppenkothen, DIRAC
  66. Visualization is about managing the viewer’s attention! Daniela Huppenkothen, DIRAC

  67. Daniela Huppenkothen, DIRAC

  68. human vision is not like photography! Daniela Huppenkothen, DIRAC

  69. pre-attentive task https://www.csc2.ncsu.edu/faculty/healey/PP/ Daniela Huppenkothen, DIRAC

  70. pre-attentive task https://www.csc2.ncsu.edu/faculty/healey/PP/ Daniela Huppenkothen, DIRAC

  71. serial search https://www.csc2.ncsu.edu/faculty/healey/PP/ Daniela Huppenkothen, DIRAC

  72. brief vocabulary lesson Semiotics: the study of signs and symbols

    in communication Everything must have meaning jradavenport
  73. time height tall short birth death Everything must have meaning

    jradavenport
  74. time height tall short birth death Everything must have meaning

    possible meanings: growth over time, growth-spurts, jradavenport
  75. time height tall short birth death Everything must have meaning

    jradavenport
  76. time height tall short birth death Everything must have meaning

    possible meanings: …tall people die? Astronomers are notoriously bad at this! jradavenport
  77. Use value-added meaning when possible! Point #5 - Repeat layouts/designs

    - give colors/shapes meanings - re-use colors consistently jradavenport
  78. If they’re thinking about your plot, 
 they’re not thinking

    about your science Point #5 restated jradavenport
  79. If they’re thinking about your plot, 
 they’re not thinking

    about your science Try to use good design principles, 
 especially contrast. Don’t tool shame Make multiple versions of a plot! Your plots (need to) tell the story at a glance! CONCLUSIONS jradavenport
  80. Extra Slides jradavenport

  81. Let’s play a game... Art or Data?

  82. None
  83. Data Art 4 hours using Eclipse, traced with IOGraph “14

    Billions” (2010) by Tomás Saraceno
  84. Art “14 Billions” (2010) by Tomás Saraceno

  85. Data 4 hours using Eclipse, traced with IOGraph

  86. None
  87. Ellsworth Kelly Jaz Parkinson Data Art

  88. None
  89. Ellsworth Kelly Data Art James Davenport