Line graphs give shape to a series of connected data values. Accordingly, they are best used when the focus of your data message is an overall pattern of data values, rather than individual values.

Tip. Lines should only be used to connect data values that are related in some way – most commonly, data for the same measure over consecutive time periods (i.e. time-series data) or a distribution of values across the range of a single measure (i.e. distribution data).

Visualising change over time for extended time-series data

Line graphs are recommended for extended time-series data (about 8 or more time points), where you want to convey change (or lack of change) in a measure over time – for example, changes in annual rainfall measurements for 30 years. Changes include trends, increases, decreases, growth, decline and rise over time.

The data in line graphs may be absolute values, percentages, proportions, ratios or rates. Individual data values for each time point are simply connected to form a line. Lines always run horizontally, left to right, to draw on readers’ instinct to perceive time horizontally:

Visualising change over time for extended time-series data on mean monthly maximum temperature in Canberra 2000 and 2015 using a line graph.

There are 2 exceptions to the use of line graphs for extended time-series data:

  • When there are fewer than about 8 intervals of measurement, consider using a vertical bar graph for time-series data. This is because lines imply trends in data over time, and the accuracy of trends is diminished with few data points.
  • When you want to emphasise specific data points or time intervals, consider displaying data in a vertical bar graph (Few 2012). This is because individual data values can be difficult for readers to identify and judge along a line.

Visualising frequency for multiple groups: frequency polygons

Like histograms, frequency polygons display the frequency or ‘counts’ of a measure across each of its possible values or intervals. Unlike histograms, frequency polygons present these data as a line of connected frequency values (i.e. a line graph), rather than individual bars.

Polygons are used instead of histograms when the measure has a large number of intervals and you want to convey a message about the shape of the overall distribution, rather than individual data points. Polygons are also useful for presenting multiple distributions (i.e. a distribution for multiple groups) on the same graph: