Data Presentation Guidelines

Table Presentation

• Round data to two statistically significant or effective numbers.

• Using three or more significant figures interferes with comparison and comprehension.

• More precision is usually not needed for epidemiologic purposes.

• Effective figures refers to numbers that contain additional, leading non-zero digits that do not vary (e.g., 123, 145, 168, or 177) or vary slightly within a column or row.

• Provide marginal averages, rates, totals, or other summary statistics for rows and columns whenever possible.

• Use columns for most crucial data comparisons.

• Numbers are more easily compared down a column than across a row.

• Organize data by magnitude (sort) across rows and down columns.

• Use the most important epidemiologic features on which to sort the data.

• Organizing data columns and rows by the magnitude of the marginal summary statistics is often helpful.

• When the row or column headings are numeric (e.g., age groups), they should govern the order of the data.

• Use the table layout to guide the eye. For example,

• Align columns of numbers on the decimal point (or ones column).

• Place numbers close together, which might require using abbreviations in column headings.

• Avoid using dividing lines, grids, and other embellishments within the data space.

• Use alternating light shading of rows to assist readers in following data across a table.

Graphical Data Presentation

• Take care in selecting a graph type in computer graphics programs. In Microsoft Excel (Microsoft Corporation, Redmond, WA), for example, you should use “scatter,” not “line” to produce numerically scaled line graphs.

• Adhere to mathematical principles in plotting data and scaling axes.

• On an arithmetic scale, represent equal numerical units with equal distances on an axis.

• When using transformed data (e.g., logarithmic, normalized, or ranked), represent equal units of the transformed data with equal distances on the axis.

• Represent dependent variables on the vertical scale and independent variables on the horizontal scale.

• Use alternatives to joining data points with a line. Consider instead

  • No line at all (use data markers only).

  • A trend line of best fit underlying the data markers.

  • A moving average line underlying the data markers.

• Aspect ratios (data space width to height) of approximately 2:1 work well. Extreme aspect ratios distort data.

• Scale the graph to fill the data space and to improve resolution. If this means that you must exclude the zero level, exclude it, but note for the reader that this has been done.

• Do not insist on a zero level unless it is an integral feature of the data (e.g., an endpoint).

• Use graphic designs that reveal the data from the broad overview to the fine detail.

• To compare two lines, plot their difference directly.

• Use visually prominent symbols to plot and emphasize the data.

• Make sure overlapping plotting symbols are distinguishable.

• When two or more data sets are plotted in the same data space

  • Design point markers and lines for visual discrimination; and

  • Differentiate them with labels, legends, or keys.

• To avoid clutter and maintain undistorted comparisons, consider using two or more separate panels for different strata on the same graph.

• When comparing two graphs of the same dependent variable, use scaling that improves comparison and resolution.

• Clearly indicate scale divisions and scaling units.

• Minimize frames, gridlines, and tick marks (6–10/axis is sufficient) to avoid interference with the data.

• Use six or fewer tick mark labels on the axes. More than that becomes confusing clutter.

• Keep keys, legends, markers, and other annotations out of the data space. Instead, put them just outside the data region.

• Proofread your graphs.

Chart Data Presentation

• Charts present statistical information comparing numeric values for sets of multiple nominative characteristics or grouped numeric characteristics.

• Data presentation is interchangeable with tables. The choice between tables and charts depends on the purpose, the audience, and the complexity of the data.

• The best charts for quick and accurate understanding are dot plots, box-and-whisker plots, and simple bar charts.

• Avoid pie charts, cluster bar charts, stacked bar charts, and other types not presented in this chapter.

• Dot plots, box plots, and bar charts are easier to understand and read if aligned horizontally (with the numeric axis horizontal).

• Sorting nominative categories by the magnitude of the numeric value helps the reader’s understanding. If the classification variable is numeric (e.g., age group), sort by the numeric category.

• The dot chart is the most versatile and the easier to understand, particularly as categories increase in number.

• Dot and box-and-whisker charts are plotted against a numeric scale and thus do not need a zero level.

• Bar charts usually need a zero level because viewers judge magnitude by the length of the bar.