Before constructing a statistical graph, an author should make two decisions. The first decision is to determine what the message is that the graph will communicate (or the purpose). The second decision is closely related to the first. The author needs to decide who the audience is and what they will expect or extract from the graph. When these decisions are made, the question of format and design can be answered. Table 1 illustrates this process. For example, if the author has a time series data set and wants to communicate variation over a period of years, he or she can choose a line chart or, for a short time series, a bar chart.

Table 1. Type and Function of Graph Message Determines Format

Thus, the fundamental task in designing a statistical graph is to focus readers' attention on the graph's data and its message, not on its design. Good design supports the data rather than the data supporting the design. A well-designed graph displays the minimum design and maximum data. A "pretty" design may attract readers' attention, but leave them confused as to what message the data are communicating. Such designs often are not presenting data clearly and accurately; they hide and distort them.   For this reason, the emphasis throughout the Guidelines is on choosing the most appropriate format or design to display the features of interest, the message, in a data set. This often can be done, as Table 1 illustrates, in more than one way. EIA products present complex data and complex messages, and the approaches to resolving format and design problems for EIA data need to be flexible. Therefore, the examples presented are not inflexible specifications but, rather, guides to methods to present data clearly and accurately. The Guidelines outlines concepts of good graphic design. It does not, except where specified in accordance with the "EIA Standard on Statistical Graphs," outline "hard-and-fast" rules.

Because it outlines concepts of good graphs design, the Guidelines summarizes and reflects the current research on visual perception for EIA authors to help them in constructing good statistical graphs. This research is an evolving field. Thus, when it changes, the Guidelines will be updated to reflect those changes.

The Guidelines does not even begin to discuss all of the formats in which to present data effectively. The "remarkable history" of information design has accumulated a rich inventory of good design.[1] Further, researchers, including those in the Federal Government, are continuously designing new formats to display data clearly and accurately.

All graphs presented in the paper version of the Guidelines were produced with EIA-supported graphics software. Unfortunately, some of the default options in graphics software are not optimal. The placing of tick marks, labels, and data points may be undesirable. Scales may not start at zero and have no indication to alert the reader. The Guidelines includes many features. Examples include break marks on the axis (as required for some graphs), clear data labels, vertical grid lines to help emphasize seasonal patterns, and harmonious shading pattern arrangements.

Changing defaults often takes time. Using templates can save time and authors are encouraged to use them, particularly for graphs that appear in recurring publications. With templates, authors only need to make the settings once and "create" a new chart - for future use - by inserting different data or making only minor alterations. [2]

The appendix of this report presents URLs where you can find additional examples and information on graph formats and current research on statistical graphs. Also provided is a glossary of terms that are used to discuss the statistical properties of graphs.

Click here to return to front of report.