Deux approches de principes à la datavisualisation

Par défaut

• Colors should be few, easily-distinguished, & meaningful
• Use direct labels, not legends
• Map data to an easy-to-compare perception (length > angle)
• Show comparisons directly
• Rank informatively

Map each data variable to a perceptual variable or aes (position, length, color…),
transform it with a stat (identity, bin, smooth, quantile…),
plot it as a geom (point, line, bar…),
and optionally facet into subplots by one or two data variables

Article disponible sur le blog Civil Statistician

Lien vers les slides