Color Maps & Palettes

Depending on your axis type you need one of two basic tools:

Continuous or practically continuous (e.g. whole numbers with high variance)

A colormap, i.e. a function mapping values to colors.

In R

Use scales or use ggplot in the first place.

In matplotlib

plt.rcParams['image.cmap'] = ..., plot(..., cmap=...)

Few discrete values (e.g. qualitative)

A palette, i.e. a fixed-size set of colors.

In R

palette() to get and palette(colors) to set.

In matplotlib

plt.rcParams['axes.color_cycle'] = ..., ax.set_color_cycle(...)

This means that you want:

Never

jet, hot, … (no perceptive continuity/uniformity, ugly, false (artifact) features, unsuited for color blindness, no guaranteed contrast)

For continuums: Perceptually uniform colormaps

• cmocean
• cubehelix (see Cubehelix.make for parametric version)

  One variant built into matplotlib.
  Parametric versions for R in destiny and rje
• viridis/plasma/magma/inferno

  Included in matplotlib. For R, use viridis

For discrete values: Fixed-size palettes with high contrast

• i want hue
• colorbrewer qualitative

  Included in matplotlib. For R, use
  scales::brewer_pal (more versatile), or
  RColorBrewer (more specific)