How to Lie with Visual Statistics

Learning how visual displays of statistics can mislead or lie will empower you to spot lies for yourself. Visual displays of data are easily manipulated, so it is important to pay close attention to a graph’s details to make sure they aren’t conveying false information.

Common Types of Graphs

When developing graphing skills, it is important to begin with the basics. Several types of graphs are commonly used by social scientists. Scatterplots depict the relation between two scale variables. They are useful when determining whether the relation between the variables is linear or nonlinear. A range-frame is a variant of a scatterplot; it provides more information with less ink by eliminating the axes below the minimum value and above the maximum value. Some line graphs expand on scatterplots by including a line of best fit. Others, called time plots or time series plots, show the change in a scale variable over time.

Bar graphs are used to compare two or more categories of a nominal or ordinal independent variable with respect to a scale dependent variable. A bar graph on which the levels of the independent variable are organized from the highest bar to the lowest bar, called a Pareto chart, allows for easy comparison of levels. Pictorial graphs are like bar graphs except that pictures are used in place of bars. Pie charts are used to depict proportions or percentages on one nominal or ordinal variable with just a few levels. Because both pictorial graphs and pie charts are frequently constructed in a misleading way or are misperceived, bar graphs are almost always preferred to pictorial graphs and pie charts.

How to Build a Graph

We first decide which type of graph to create by examining the independent and dependent variables and by identifying each as nominal, ordinal, or scale. We then consider a number of guidelines to develop a clear, persuasive graph. It is important that all graphs be labeled appropriately and given a title that allows the graph to tell its story without additional text. For an unambiguous graph, it is imperative that graph creators avoid chartjunk: unnecessary information, such as moiré vibrations, grids, and ducks, that clutters a graph and makes it difficult to interpret. When using software to create graphs, it is important to question the defaults built into the software and to override them when necessary to adhere to these guidelines.

Finally, keeping an eye to the future of graphing—including interactive graphs, computer-generated maps, and multivariable graphs—helps us stay at the forefront of graph making in the behavioral sciences.