Distribution charts

Distribution charts show how the values of a continuous numeric variable are spread and where they cluster, rather than only totals or averages.

Chart Shows Best for
Histogram Counts per value range (bin) The overall shape of one continuous variable
Density plot A smoothed histogram Comparing the shapes of two or three groups
Box plot Median, quartiles and outliers Comparing many groups compactly
Violin plot A box plot with density shape Many groups, when shape matters
Histogram of households by distance to the nearest health facility, in 2 km bins
Box plots of clinic waiting times for four regions, sorted by median
Density plot of clinic waiting times in two regions: a tall narrow peak for South, and a lower, wider curve with two peaks for East
Violin plots of clinic waiting times for four regions, sorted by median, with the median as a white line
  • Avoid them for audiences unfamiliar with statistical charts (for them, describe the distribution in words or use a simple histogram), for very small datasets, and when exact values matter more than patterns.
  • With too many bins, a histogram shows noise. With too few, it hides patterns. Try several bin sizes, and pick round-number bin edges.
  • Decide deliberately whether to explain, transform or show outliers separately. Don’t just let the axis stretch.
  • A histogram is also the quickest way to choose map classes.