Common issues in tables and figures

Please ensure your figures and tables adhere to the guidance provided below. See also this series of videos from Maarten Boers on improving figures and tables.

Checklist for tables

  • Reduce precision to integer. General rule for precision: Usually a maximum of 2, and rarely 3 or more significant digits, regardless of source data precision. In many cases (e.g., percentages), integers suffice!
    • A significant digit is one that carries information. Results are rounded to the nearest significant digit. In case of 5, to the nearest even digit.
    • This applies to all statistical parameters: — mean, standard deviation, median, min, max, percentiles etc — and all percentages.
    • Going to 3 significant digits might be required for some primary outcome (usually specific efficacy) measurement instruments, and in cases where results range between powers of ten (e.g. 0.50–9.00, or 35–128). Two significant digits would be enough for almost all other measures.
  • Alignment:
    • Named categories, dates: left
    • Numbers: right, and on decimal point (if applicable)
    • Multiple entries in one cell: specific advice
  • Logical order of categories (rows) and results (columns)

Checklist for figures

  • General
    • Increase resolution (minimum 300 dpi)
    • Delete background color/shading
    • Increase font size of tick labels, axis titles
    • Avoid stacked bars, pies, 3D depiction of 2D data
    • Use bar graphs only for results that are natural quantities (e.g. counts, percentages of a whole; but not for mean concentrations)
    • Depict low numbers of observations as individual symbols, not (only) as summary measure
    • Depict as many summary measures as possible (e.g. box plot better than mean ± error bar)
  • Data depiction
    • Increase thickness of data series
    • Do not use dashed lines for data series. Use greyscale or color tones instead
    • Do not use patterns for shading. Use greyscale or color tones instead
    • Increase size of data symbols
    • Change data symbols:
      • Delete outline around bars
      • Decrease thickness of error bars
      • Delete horizontal whiskers on error bars
  • Axes
    • Decrease thickness of axis lines and ticks
    • Put y axis title horizontal, left aligned above y axis or centered above panel
    • Scale: optimize scale for spread of data (i.e. try to prevent large areas of white space), but try to be consistent across figures
    • Axis offset: if necessary, use offset to avoid axis lines superimposed on data
    • Axis break: best avoided. Never use with bar graphs. If unavoidable, the break should be through the whole figure (e.g. no lines crossing the break) and clearly marked on the axis. Use a scale marker (eg, a small bar) if the scale differs before and after the break.
  • Matrix (multipanel) plot:
    • Move panels very close together;
    • For both the group of y axes and also the group of x axes, strive to make the axis scale (numerical range, or categories) the same. If this is possible, the tick labels of the inner panels can be removed:
      • For panels in columns: delete x axis title and tick labels (keep ticks) of all except bottom panel, and move panels very close together.
      • For panels in rows: delete all y axis titles and tick labels, put y axis title horizontal left aligned above y axis, or centered above top panel
  • In-figure legend:
    • Avoid if possible or place well out of the way of the data. Alternatives: place in caption text, or use labels close to series.
    • If essential and placed outside data frame, on top or below is better than to the right (graph will be downsized to fully fit on page).