Figure and Table Reporting: Standards for Data Visualization

No. Present each data point in one place only. In the text, highlight key findings and direct readers to the table or figure (e.g., 'Treatment improved symptoms in 85% of participants (Table 1)'). Do not repeat the table values in the narrative text; it wastes space and risks introducing transcription errors.

A caption should be a complete explanation that allows readers to understand the figure without consulting the text. Begin with a brief title (e.g., 'Survival Curves by Treatment Group'), then explain what is shown, the sample size, and any important details (e.g., methods, definitions of outcomes). If the figure uses symbols or error bars, explain them in the caption or a legend. Length: typically 1–3 sentences, but more if necessary for clarity.

Yes, but use color wisely. Use color to highlight important patterns or to distinguish groups, not for decoration. Ensure that figures are intelligible in grayscale (print some figures in black-and-white to test). Avoid red-green color combinations (color-blind readers cannot distinguish them). Use colorblind-friendly palettes (available online, e.g., ColorBrewer). Label colors or use symbols in addition to color (e.g., solid line vs. dashed line for different groups).

A figure is a complete graphic (one chart, image, or diagram). A figure panel is a component of a larger figure. For example, 'Figure 1' might contain three panels (A, B, C), each showing a different aspect of the data. This allows you to present related data as one cohesive figure while maintaining clarity. Label panels as 'Figure 1A,' 'Figure 1B,' etc.

Report values with appropriate precision based on the measurement and the sample size. Generally, do not report more than one extra significant figure beyond what the data support. For example, a mean with a sample size of 20 should typically be reported to 2–3 significant figures (e.g., 45.2), not five (45.234). Standard deviations and error bars should use one more significant figure than the mean (e.g., mean 45.2, SD 8.3). Check your field's conventions; some fields are more conservative (fewer figures), others more liberal.

Use error bars for continuous data with variation (means with standard deviations or standard errors, or confidence intervals). Do not use them for categorical proportions without context. In figure captions, always specify what error bars represent (e.g., 'bars represent ±1 SD' or 'error bars are 95% confidence intervals'). In bar graphs comparing groups, confidence intervals are often preferred over standard errors because they aid interpretation (non-overlapping CIs suggest significant differences, though this is not a formal test).