Most barcharts are graphical representations of frequency tables and crosstabulations, i.e. they let you visually examine the distribution of counts and percentages. Instead of showing s relative (percentage) or absolute (count) of a categorical variable, a bar can also represent a statistic of another variable, e.g. averages of continuous variables or percentages below or above some threshold.

There are several ways to produce barcharts in SPSS:

• Use (details below) or if you want three dimensional barcharts
• Some statistical procedures have options to produce straighforward barcharts:
  • Frequencies (Chart button on the dialog panel) produces a simple barchart (option to depict counts or percentages9.
  • Crosstabs (check box on the main dialog) produce a simple bivariate barchart showing counts only (no options)
  like frequencies (Chart button on the dialog panel) and Crosstabs (check box on the main dialog) have options to produce straightforward barcharts
• Barcharts produced with the (details below)
• Barcharts from pivot tables. [See below for details]

will ask you first what kind of barchart you would like to produce. According to your selection a number of different types will be produced: barcharts depicting a single variable, barcharts showing two variables, either side by side (clustered, dodged) or one on top of the other (stacked); you can also compare several variables. With all combinations of options, a wide variety of barcharts can be produced. Please understand that not all forms of barcharts are meaningful for all situations; the default SPSS proposes corresponds to the most common use but may not be what you want.

This dialog comes up with the default selection.

By default a barchart shows the number of observations in each bar; select % of cases to show percentages; Instead of absolute or relative frequencies of the variable depicted, a statistic for another variable can be shown instead.

If you select Other Statistic you will have to select an additional variable; by default the mean is used, e.g. you might show the average age for every category of the "Satisfaction of life" variable. You can select another statistic using Change Statistic (see on the left).

You should however be very careful to select the appropriate statistic to show, as for instance the mean of a categorical variable does not make sense. Although the mean makes sense for a continuous variable, e.g. age in years, charts with bars showing just the mean for each category are rather poor in content, a much better alternative would be a boxplot.

If you select Clustered or Stacked on the inital dialog, a second categorical variable can be defined as illustrated by the example (clustered barchart).

Standard barcharts are used to depict a single (categorical variable) and other forms (stacked, clustered) allow for the comparison of several variables. In addition to these possibilities, barcharts can be arranged in "panels", i.e. barcharts produced for all categories of one or more other categorical variables. Here's an example depicting "Satisfaction with Life", across categories of "sex" and "religion", i.e. in the top left corner you see a barchart of satisfaction with life for male catholics.

Here's the dialog that lets you produce this panel plot. Note by the way that panel plots can be obtained for all graphic types.

• See also Percentages in bar charts for more complex barcharts

Barcharts can also be produced from numbers in tables in the .

The example illustrates a barchart produced from the selected percentages in a frequency table. After selecting the data you wish to depict, choose from the context menu. Make sure to select only the percentages of interest, i.e. in the example, do not select the 100%, as you do not want a misleading total bar to dominate the chart. (If you select it you can of course remove it, using the , but then you would also have to adjust the scales of the chart).

This works from any pivot table. Be very careful when selecting the numbers/cells to be shown in the chart, and verify the resulting graph, as it might need some ajustments (elements shown, scales and especially legends, titles etc.)

Like all charts, a barchart can be edited using the .