Which graph should I use?

This display, in combination with well-crafted words, is a comprehensive and powerful tool for scientific communication.

Data visualisations are different from technical infographics, which represent a wider range of information and concepts.

Choosing the best graph for your data

Graphs are the most common and recognisable way to visualise data. When you choose the right graph, your readers will easily and accurately understand what your data show. Download our factsheet: What type of graph is best for my data? [994KB].

Some quick guidance is also provided for different data stories here.

Data story: Compare data values across independent items, groups or categories.

• Recommended graph type: Horizontal bar graph.
• Other graphs: Dot graph.
• Tips:
  • Order bars by size to emphasise differences.
  • Use clustered bars for groups with subcategories.

Data story: Show how a measurement changes over several time points.

• Recommended graph types: Line graph for many time points; vertical bar graph for few time points.
• Other graphs: Dot graph; dumbbell graph for 2 time points for several groups.
• Tips:
  • Only connect consecutive data values with lines.
  • Lines always follow a horizontal direction from left to right, with time intervals on the x axis.

Data story: Show how data values relate to, compare with, or make up a total measurement.

• Recommended graph types: Horizontal bar graph for a single population (bars represent parts or proportions); stacked bar graph for proportions of a measure across multiple populations.
• Other graphs: Vertical stacked bar graph for proportions of a measure across a small number of time points; stacked area graph for proportions of a measure across many time points.
• Tips:
  • Parts must add to 100% if they represent percentages, or the total absolute value for other measures.

Data story: Show how frequency or count values are distributed over the range of a single measure.

• Recommended graph types: Vertical bar graph (histogram) for measures with a small range of possible values; line graph (frequency polygon) for measures with a large range of values.
• Other graphs: Horizontal strip plots and box plots.
• Tips:
  • Strip plots and box plots may be unfamiliar to some readers – consider using a simple line or bar graph instead.

Data stories: Show the spread of data values on a single measure over time, or show how data values are distributed for several categories or groups at a single time point.

• Recommended graph types: To show the spread of data values on a single measure over time, use a line graph with shaded lower and upper bounds that represent this spread.
• To show how data values are distributed for several groups at a single time point, use vertical strip plots, box plots and trellises (panels) of bar or line graphs.
• Tips:
  • Measures of error around data values are often better understood as shaded bounds rather than vertical lines (error bars).
  • Strip plots and box plots may be unfamiliar to some readers – consider using a trellis of bar or line graphs instead.

Data story: Show the difference between data values and a baseline, target or threshold.

• Recommended graph types: Vertical bar graph with a reference line to show differences between values for categories or groups and that reference; line graph with reference line to show differences between data values and that line over time.
• Tips:
  • Data values above the reference indicate positive differences; data values below indicate negative differences.
  • The y axis can measure absolute differences or percentage change between data values and the reference.

Data story: Show an association between 2 measures or variables.

• Recommended graph type: Scatter plot.
• Other graphs: Side-by-side horizontal bar graphs to show an association between 2 measures when scatter plots are unfamiliar to readers (most effective for linear associations).
• Tips:
  • Consider including a trend line on scatter plots to highlight the type and strength of association.

Data story: Compare the amount of difference between 2 data points across multiple groups. These data may be the same measure over 2 time points (eg pre- and post-tests), or values of the same measure for 2 related subgroups (eg income for urban and rural cohorts across several locations).

• Recommended graph types: Dumbbell graph or clustered bar graph.
• Tips:
  • It is difficult to compare the amount of difference between clustered bars for a large number of groups because of the distance between first and last groups along the x axis.
  • Dumbbell graphs may be less familiar than clustered bars to some readers.