Share
Embed
Transcript
25 slides
-
Small Team, Big Value Using R to Design Visualizations
Ian Lyttle, Schneider Electric @ijlyttle -
Who are Schneider Electric? • We are more than 150,000
employees worldwide. • We help our customers use electricity safely and efficiently. -
How do we use data? Our products collect and transmit
data on things like: - power consumption - number of operations - room temperature Our customers have questions like: - “When will I need to replace the power-supply battery?” - “Did my facility use too much energy last month?” - “Do I have enough cooling in my data-center?” -
SE businesses use different visualization toolsets, including: Each runs in
the browser and uses JavaScript. None of these is ggplot2. SE Customer-Facing Tools -
SE businesses each have data-science teams. SE also has a
central data-science team. This is who I work with. Import Tidy Transform Visualize Model Communicate Wickham & Grolemund, R for Data Science How are we organized? -
How are we organized? Import Tidy Transform Visualize Model Communicate
Wickham & Grolemund, R for Data Science Import, Tidy, Transform, Model -Larger team, based in France -Uses Python -Deploy models and data as web services https://exchange.se.com/develop Visualize, Communicate -Smaller team, based in US -Uses R -Deploy practices -
How are we organized? Import Tidy Transform Visualize Model Communicate
Wickham & Grolemund, R for Data Science Visualize, Communicate -Smaller team, based in US -Uses R -Deploy practices Ian Lyttle Haley Jeppson PhD Student, Iowa State University -
By Mike Murphy, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=295635 Data-Visualization Guidelines for
Customer-Facing Tools Challenge from SE UI Design Team: -
• Solid advice • Uses ggplot2 • No code in
book Claus Wilke Technique https://serialmentor.com/dataviz/ -
Jenny Bryan Tooling https://twitter.com/JennyBryan/status/878016566826225664 • Focus on what’s important. •
Work towards good-enough. • Workflow to support this: • {usethis} package • https://happygitwithr.com • https://rstats.wtf -
By Libby VanderPloeg, https://giphy.com/gifs/cute-feminist-girlpower-3o7abBphHJngINCHio Community Gabriela de Queiroz • Founder,
R-Ladies • Began in San Francisco • Now hundreds of chapters • Diversity of input leads to better outcomes. -
SE Categorical Palette SE brand colors optimized for marketing: Not
optimized for data-visualization. • Propose a related categorical-palette • Take into account color-vision deficiency • Remain “Schneidery” Challenge -
SE Categorical Palette SE brand colors optimized for marketing: Not
optimized for data-visualization. • Propose a related categorical-palette • Take into account color-vision deficiency • Remain “Schneidery” Challenge Technique • Book offers useful advice • {colorspace} package • Achim Zeileis @ UseR! 2019 • Think in terms of hue, chroma, luminance -
Hue Chroma Luminance Color-vision deficiency Perceptual distance Visible Gamut RGB
Color space Just-noticeable difference CIEDE-2000 “Crayola bright” -
SE Categorical Palette SE brand colors optimized for marketing: Not
optimized for color-vision deficiency. • Propose a new palette • Take into account color-vision deficiency • Remain “Schneidery” Challenge Technique • Book offers some advice • {colorspace} package • Achim Zeileis @ UseR! 2019 • Think in terms of hue, chroma, luminance -
SE Categorical Palette Technique • Book offers some advice •
{colorspace} package • Achim Zeileis @ UseR! 2019 • Think in terms of hue, chroma, luminance • Need numbers to show perceptual-differences • {farver} package • {paleval} ijlyttle/paleval: • Built on {colorspace} and {farver} • Good enough for now, improve later Tooling -
SE Categorical Palette • Need numbers to show perceptual-differences •
{farver} package • {paleval} ijlyttle/paleval: • Built on {colorspace} and {farver} • Good enough for now, improve later Tooling • Created internal blog • {distill} via GitHub Enterprise Communication -
SE Categorical Palette • Seek more opinions, new perspectives •
Community@Work, Data Visualization: • By practitioners for practitioners • Point-of-contact with SE UI Design Team Community • Created internal blog • {distill} via GitHub Enterprise Communication -
SE Categorical Palette Proposed categorical palette: • Takes into account
color-vision deficiency. • Potentially “Schneidery”. • Useful feedback: may remove reds. Designed using R, delivered as hex-codes. Result -
More Broadly Grammar-of-graphics as description and tool (ggplot2) Interactivity: Vega-Lite
vegawidget/ggvega Move things off your computer into the world! Get to know a local university -
Small Team, Big Value Using R to Design Visualizations
Ian Lyttle, Schneider Electric @ijlyttle @vegawidget, @r-box https://exchange.se.com/develop