Practical Data Visualization

Transcription

1 Practical Data Visualization March 18, 2015 COMPSCI 216: Everything Data Angela Zoss Data Visualization Coordinator Data and Visualization Services

2 WHY VISUALIZE?

3 Preserve complexity Anscombe s Quartet I II III IV x y x y x y x y

4 Preserve complexity Anscombe s Quartet I II III IV x y x y x y x y Property Mean of x 9 (exact) Variance of x 11 (exact) Value Mean of y 7.50 (to 2 decimal places) Variance of y or (to 3 decimal places) Correlation between x and y Linear regression line (to 3 decimal places) y = x (to 2 and 3 decimal places, respectively)

5 Preserve complexity Anscombe s Quartet

6 Evaluate data quality Query using Facebook API Node-link diagram Kandel, Heer, Plaisant, et al. (2011)

7 Evaluate data quality Query using Facebook API Node-link diagram Matrix display with clustering Kandel, Heer, Plaisant, et al. (2011)

8 Evaluate data quality Query using Facebook API Node-link diagram Matrix display with clustering Matrix display, API return order Kandel, Heer, Plaisant, et al. (2011)

9 Evaluate data quality Query using Facebook API Node-link diagram Matrix display with clustering Matrix display, API return order 5000-item result limit Silent failure Kandel, Heer, Plaisant, et al. (2011)

10 Tell a story Hans Rosling The River of Myths

11 CREATING A VISUALIZATION

12 From Data to Graphic What data types are present in the data source? Categorical? Numerical? Relational?

13 Matching Data Types to Visual Elements Mackinlay, J. (1986). Automating the design of graphical presentations of relational information. ACM Transactions on Graphics, 5(2),

14 From Data to Graphic What data types are present in the data source? What type of analysis do you want to support? Are you looking for correlations? Distributions?

15 Variable Width Column Chart Table or Table with Embedded Charts Bar Chart Column Chart Circular Area Chart Line Chart Column Chart Line Chart Scatter Chart Column Histogram Line Histogram Bubble Chart Scatter Chart 3D Area Chart Stacked 100% Column Chart Stacked Column Chart Stacked 100% Area Chart Stacked Area Chart Pie Chart Waterfall Chart Stacked 100% Column Chart with Subcomponents

16 From Data to Graphic What data types are present in the data source? What type of analysis do you want to support? What visualization type seems to be the best fit for the goal? Do you want the visualization to be accessible for a broad audience? Flashy and engaging? Convincing?

17 POSITION IS

18 Basic tips Rotated text is harder to read People are very good at reading x/y position, bar length People are not as good at reading angles, areas Avoid overlap by filtering, aggregating, leaving space

19 COLOR IS

20 Basic tips For categorical variables: People have trouble differentiating between more than 5-7 hues (colors) For numerical variables: People have trouble differentiating between more than 5-7 shades Rainbow color gradients are very problematic For highest contrast, only use color to highlight

21 VISUALIZATION TYPES

22 Showing Values

23 Basic charts and graphs

24 Binned Scatterplot

25 Parallel Coordinates

26 Sankey/Alluvial Diagram

27 Heat Maps

28 Pairs Plots Dynamic Pairs Plot:

29 Showing Distributions One-dimensional scatter plot Histogram

30 Showing Space

31 Proportional symbol

32 Proportional symbol

33 Choropleth

34 Choropleth

35 And don t make users do visual math.

36 Common Routes Based on Ship Log Data

37 Atlas of the Historical Geography of the United States (1932)

38 Possible tools for mapping ArcGIS QGIS Tableau Public CartoDB Google Fusion Tables Google Earth GeoCommons JavaScript D3 Leaflet Kartograph Polymaps Google Maps API developers.google.com/ maps/documentation/ javascript/ Very basic: Google Spreadsheets BatchGeo OpenHeatMap See also:

39 For congress data in Tableau congressional-districts online/en-us/help.htm#maps_geographicroles.html

40 Showing Time

41 Economic indicators over time

42 Time series of 2D data set

43 Connected Scatterplot

44 Stream graphs diseases of the circulatory system Japanese German Russian French English diseases of the digestive system endocrine, nutritional and metabolic diseases infectious and parasitic diseases injury, poisoning and other external causes mental and behavioral disorders cancer (neoplasms) pregnancy and childbirth diseases of the respiratory system

45 Storylines

46 Shape of Song

47 Over the Decades, How States Have Shifted

48 Possible tools for temporal vis. Basic charting tools Raw TimelineJS Simile Timeline D3

49 Showing Relationships

50 Edges

51 Nodes

52 Both

53 With color and size coding

54 Bipartite graph, alluvial diagram

55 Circular layout/chord diagram

56 Tube Map

57 Possible tools for network vis. D3 Gephi NodeXL Pajek networks/pajek/ Cytoscape Network Workbench/Sci VOSviewer UCINET ucinetsoftware/home GUESS R SigmaJS Circos

58 Showing Text th diseases of the respiratory system

59 Word cloud diseases of the circulatory system diseases of the digestive system endocrine, nutritional and metabolic diseases infectious and parasitic diseases injury, poisoning and other external causes mental and behavioral disorders cancer (neoplasms) pregnancy and childbirth diseases of the respiratory system

60 Bubble Plot

61 Frequencies over time

62 Scatter Plot

63 Sentiment analysis

64 Sentiment analysis

65 Word Tree

66 Word co-occurrence network

67 Phrasenet

68

69 VISUALIZING UNCERTAINTY

70 Projections

71 Missing data

72 Alternative solutions

73 Take-away Uncertainty is blue.

74 TOOLS THAT DON T NEED INSTALLATION

75 Plot.ly

76 Plot.ly Browser based (or Excel add-in) Makes wide variety of chart types Allows for python, MATLAB, R, etc. syntax Makes charts that are hosted/shareable

77 Example: Bubble chart

78 Raw Has visualizations to show: Numbers Relationships Hierarchies

79 Raw Paste in a data table (.csv,.tsv, copied from Excel) Select chart type Drag column headers to different chart attributes Save out image or SVG code

80 Example: Alluvial Diagram

81 Google Spreadsheets

82 TimelineJS

83 Timeliner

84 StoryMapJS

85 Also, GitHub auto-rendering 3D Files 3d-file-viewer GeoJSON/TopoJSON mapping-geojson-files-on-github CSV/TSV rendering-csv-and-tsv-data

86 SOFTWARE APPLICATIONS

87 JMP Pro

88 JMP: Essential Graphing Overlay Plots Scatterplot 3D Contour Plots Bubble Plots Parallel Plots Cell Plots Treemaps Scatterplot Matrix Ternary Plots Summary Charts Create Maps

89 Example: Contour Plot

90 JMP Pro Statistical software Drag-and-drop chart builder Good charting options, including a basic map Can save code for all charts (good for reproducibility) Can save vector graphics from charts (good for print publications and graphic design work)

91 Tableau

92 What can Tableau make? Text tables Heat maps a grid representing variables by size and color Highlight tables a grid representing variables by text and color Maps (symbol, filled) Pie charts Horizontal bars Stacked bars Side-by-side bars Treemap a grid representing variables by size Circle views Side-by-side circles Lines/Area charts Lines/Area charts (discrete) Dual lines Dual combination Scatter plots Histogram Box-and-whisker Gantt Bullet graphs Packed bubbles/ Word cloud

93 Tableau Desktop Built specifically for visualization Can create interactive charts and dashboards Can post to the web (but make sure data are safe to share) Not great for print charts (basically have to take screenshots) Free for students:

94 Example: Animated Map

95 Gephi

96 Data formats Confusing number of choices GEXF supports many program features, but a pain to write by hand Spreadsheet is convenient and supports important features

97 In addition to network visualization, Gephi can calculate: Degree (when directed, in-degree and out-degree) Diameter Betweenness Centrality Closeness Centrality Eccentricity Density Clustering/Modularity

98 ADVANCED TOOLS

99 D3.js

100 About D3 JavaScript library Fairly low level; building with rectangles and circles and lines, instead of pre-made chart structures* Basic functioning makes it easy to join HTML elements with data points

101 *D3 Middleware Basic line/area chart: xcharts Rickshaw (specifically for time series) NVD3 Vega ~10 lines? ~16 lines ~31 lines ~57 lines /ch02.html#_tools_built_with_d3

102 *D3 Middleware, cont d. DC ( good for dashboards (includes Crossfilter) D3plus ( good for tool tips and info panels Dimplejs ( good for annotations, very pretty

103 D3 Resources Interactive Data Visualization for the Web Tutorial and Cheat Sheet, c tutorial-at-visweek-2012/ D3 Tips and Tricks

104 When to use D3 Need for customized chart types ( Want to use JavaScript Have only a low number of data points or elements (SVG vs. HTML5 Canvas) Want to have it on your résumé

105 D3 workshop tomorrow! Visualization in d3 Thursday, March 19, 7-9pm Edge Workshop Room (Bostock 1 st Floor) (Workshop is full, but if there are no-shows you could try to sneak in.)

106 Python Bokeh web visualizations with big datasets

107 Python ggplot2 for python includes good graphical principles

108 Python Anaconda good for installing many data analysis packages, including matplotlib

109 R Shiny

110 ELK stack

111 ELK stack Elasticsearch flexible and powerful open source, distributed, real-time search and analytics engine full-text search (lucene) plus fast queries and many built-in aggregations for large data (timebased and stats w/facets) Logstash helps you take logs and other time based event data from any system and store it in a single place parse Kibana Elasticsearch s data visualization engine sharable dashborads for real-time, interactive visual exploration

112 ELK stack Open source, but company builds APIs for all major languages Potential end-to-end solution for storage, plus monitoring by both developers and customers Geared towards large time-based, geo-spatial, and textual data Free for academic use Security product is pay only

113 MORE TIPS

114 Good Chart Makeover Examples The Why Axis chart remakes Storytelling With Data visual makeovers: label/visual%20makeover

115 On the web Bad examples: WTF Viz, Good examples: Thumbs Up Viz, Ask for help: Help Me Viz,

116 More on Data Visualization Visual communication: Data visualization: Top 10 dos and don ts for charts and graphs:

117 GETTING HELP

118 Data and Visualization Services Data collections, LibGuides, etc. Blog (tutorials, announcements, etc.) Walk-in consultations (or by appointment Data and Visualization Lab in the Edge (fast hardware, diverse software) Additional workshops (listserv

119 QUESTIONS? SUGGESTIONS?