Data Viz
Drawn from Fundamentals of Data Visualization by Claus O. Wilke
Ugly, bad, and wrong figures
3D figures are confusing
Representing multi-dimensional data
Make things readable
Make things readable
Make things easy
If there is a clear visual ordering in your data, make sure to match it in the legend.
Make things easy
If there is a clear visual ordering in your data, make sure to match it in the legend.
Make things easy
Show the patterns and no extraneous information
Make things easy
Show the patterns and no extraneous information
Make things easy
Show the patterns and no extraneous information
Qualitative Color Scales
Recommended: Okabe Ito, ColorBrewer Dark2, ggplot2 hue
Qualitative Color Scales
Recommended: Colorbrewer Blues, Heat, Viridis
Diverging color scales
Recommended: CARTO Earth, ColorBrewer PiYG, Blue-Red
Adding highlights
Recommended: Okabe Ito Accent, Grays with accent, ColorBrewer Accent
Use color for a purpose
Make sure it grayscales
Don't include too much information
Use qualitative color scales for three to five different categories
Consider color deficiencies: sequential scale
A properly designed sequential scale (e.g. Heat) presents a continuous gradient from dark to light colors
Consider color deficiencies: diverging scales
Popular color contrasts in diverging scales can become indistinguishable for some forms of color-vision deficiency. ColorBrewer PiYG (pink to yellow-green) works for all forms of color-vision deficiency.
Consider color deficiencies: qualitative scales
Qualitative scales (e.g. Okabe Ito) require that many different colors are distinguishable from each other under all forms of cvd.
Visualizing uncertainty
Error bars don't represent the variation within each category or the uncertainty of the sample means well.
Visualizing uncertainty
Kernel density plots help to visualize several distributions at once.
Consider what you want to communicate
Consider what you want to communicate
