Principles of Visual Design

FOR ALL VISUALS

  1. Don’t use a visual if the information is easily conveyed in writing within the main text.
  2. Use the type of visual that presents the information most clearly.
  • Is a table or figure better?
  • If a figure, which type of figure is best?
  1. Design the visual. 
  • Decide what you are trying to communicate and craft the visual to accomplish that.
  • Don’t copy output from code/software directly into your paper. (Code output is for you, not your readers!)
  1. Use words, not variable names, whenever possible.
  2. Round all quantities appropriately. 
  • Don’t paste numerical output from code. 
  • Don’t include more digits than are meaningful for the context.
  1. Include a title and caption
  • Titles should be simple, descriptive, and specific to the content of the visual.
  • Captions should include information that can help readers understand the visual while their eyes are on it.

FOR GRAPHS

  1. Make sure everything on the graph is easy to see/read.
    •    Label axes—clearly. Use large, sans serif fonts.
  2. When feasible, avoid legends and label curves directly instead.
  3. Choose colors and symbols carefully.
  • Avoid symbols that are confusing when overlapping. 
  • Select colors intentionally. Use palettes that are color-blind-friendly and distinguishable when printed in grayscale.
  • Use redundant indicators to differentiate different groups of data (e.g., both color and shape).

FOR TABLES

  1. Think carefully about what quantities to present to readers—given what you want them to see.
    •    For example, functions of parameters may be more important than the values of parameters themselves (particularly for interactions and polynomials)
  2. Choose the order of rows and columns carefully.
    •    Is there a conventional order that readers will expect? If yes, stick to it.
    •    If not, consider natural ordering of numerical values or placing the most important information on top.
  3. Don’t use more rows or columns than necessary. 
  • For example: keep all parts of a confidence interval together in one column.