Components of Figures
Components of Figures
Components of Figures
Clear display of data or information is the most important aspect of any figure. For figures that display quantitative information, data values may be represented by dots, lines, curves, area, length, or shading, based on the type of graph used.
Scales for Graphs.
The horizontal scale (x-axis) and the vertical scale (y-axis) indicate the values of the data plotted in a graph. In most graphs, values increase from left to right (on the x-axis) and from bottom to top (on the y-axis).
Range of Values.
The range of values on the axes should be slightly greater than the range of values being plotted, so that the entire data set can appear within the area defined by the axes and most of the possible range of values on the axes will be used. Ideally, the range should include 0 on both axes, if 0 is a possible value for the variable being plotted. In line graphs, if a large range of values is necessary but cannot be depicted with a continuous scale, discontinuity in the axis should be indicated with paired diagonal lines that signify a missing portion of the range (//).15 Numerical data on 2 sides of a scale break should not be connected to avoid the implication that data on either side of the discontinuity are linear. For single-axis plots, data that exceed the limits of the axes can be indicated with an arrowhead.
Divisions of the scales on the graph axes should be indicated by intervals chosen to be appropriate, simple multiples of the quantity plotted, such as multiples of 2, 5, or 10.15 Numbers that represent the values on the axis scale are centered on their respective tick marks. For linear scales, the axis must appear linear, with equal intervals and equal spacing between tick marks. However, logarithmic scales may be useful to show proportional rates of change (Example F13) and to emphasize the change rate rather than the absolute amount of change when absolute values or baseline values for data series vary greatly.
Axes should be labeled with the type of data plotted and the unit of measure used. Data may represent numerical values, percentages, or rates. For numerical data, customary units of measure and their respective abbreviations or symbols should be used (see 14.12, Abbreviations, Units of Measure). In single-axis graphs, categories should be clearly labeled along the baseline.
Symbols, Patterns, Colors, and Shading.
Symbols, line styles, colors, and shading characteristics used in the figure must be explained, preferably by direct labeling of components in the figure or in a key. Alternatively, this information may be included in the legend. For a series of figures within an article, the types of symbols, line styles, colors, and shading should be used consistently. For example, if data for the intervention group and for the control group are designated as a heavy line and as a lighter line, respectively, then these same line styles should be used for similar data for these groups in subsequent figures.
When data points are plotted, symbols should be distinguished easily by shape and color or shade. For example, if 2 symbols are needed, the recommended symbols are ○ and ●,15 although ☐ and ▀ or △ and ▴ may be used. A combination of these symbols can be used when 3 or more symbols are required. The shading or color of the symbols can designate specific data. For instance, in all figures in an article, ○ may indicate data for the placebo group and ● for the intervention group.
In bar charts and other figures (such as maps), shading is preferable to cross-hatching and other patterns to distinguish groups. Patterns can be difficult to read both in print and online.5 Shades should be of appropriate gradations to show contrast (eg, 10%, 40%, and 70% black).
Box and Whisker Plots.
Box and whisker formatting may be useful to illustrate the nonnormal distribution of values within a group (data set). Typically, the top and bottom of the box represent the 25th and 75th percentiles, the horizontal line inside the box represents the median or mean, the whiskers are the 10th and 90th percentiles, and any outliers are shown as circles (Example F11). Because the value of each of these components may vary, it is important to define them. Mean values in box and whisker plots may be connected by curves to show trends, such as point estimates of mean values.
For plotted data, error bars (depicting standard deviation, standard error, range, interquartile range, or confidence intervals) are an efficient way to display variability in the data.16 Error bars should be drawn to encompass the entire range of variability, not in just one direction (Example F7). Error bars should always be defined either in the legend or on the plot itself.
In most cases, figures should not be presented in 3-dimensional format, even when data for 3 variables are being displayed. A 3-dimensional presentation is inappropriate for any figures that contain only 2 dimensions of data. Many software programs allow users to add enhancing elements to figures, but 3-dimensional display may confuse readers or distract from important graphical relationships. For instance, it may be difficult to read from the bar to the correct value on the axis. Most 3-dimensional presentations can be replotted into more straightforward graphics (Example F21, Example F22).