Visualization
Basic principles of confirmatory visualization
- Show the design — plot the experimental structure (conditions, factors) so the reader sees what was manipulated.
- Facilitate comparison — put the quantities to be compared adjacent or overlaid; don't make the eye do subtraction.
- Show the data — points or distributions behind summaries reveal spread that a bar alone hides.
- Maximize information, minimize ink — every pixel should earn its place (Tufte's data–ink ratio).
Visualizing uncertainty — error bars (standard error or confidence interval) are mandatory for estimation-based reporting (see Estimation and Inference).
Exploratory visualization
Before and during analysis, plot to learn:
- Distributional information — histograms and densities expose skew, outliers, and the "gorillas" you'd miss in a mean.
- Data diagnostics — missingness, outliers, and impossible values.
- Plot data-collection details — time of day, order effects, duplicate IDs.
- Hypothesis-driven exploration — check specific predictions informally.
- Visualization as debugging — a weird plot often reveals a coding or labeling bug.
Key terms
- Confirmatory visualization — plots that communicate a pre-specified result.
- Exploratory visualization — plots used to discover patterns and problems.
- Data–ink ratio — the share of ink that represents data, not decoration.
Further reading
- Previous: Writing · Next: Meta-analysis · Hub: Experimentology
- The estimates you plot: Estimation
- Cognition Psychology — effect plots for cognitive paradigms.
- Stroop Effect — a result worth plotting with error bars.
- Plot boxplots instantly from your own data: Boxplot Calculator