statistical-and-uncertainty-visualization

Match the chart to the statistical question: histograms show distribution shape, box plots summarize spread and outliers, and violin or density plots reveal multimodality. Use summaries only when they do not hide the variation that matters to the decision.

Use error bars, error bands, intervals, or ensemble views depending on whether the uncertainty comes from estimation, simulation, or sampling variability. Always define what the uncertainty measure actually represents so viewers can interpret it correctly.

Pair intervals with a plain-language explanation of what they mean and how much they change the conclusion. Avoid decorative uncertainty bands, and state the interval definition explicitly rather than assuming it is self-explanatory.

Missingness deserves explicit visual treatment when it affects trust or comparability, such as annotations or dedicated marks. Never encode missing values as zero, since that silently distorts the analysis.

Aggregation, smoothing, truncation, and selective baselines can distort conclusions even when rendering is clean. Comparing groups with different sample sizes as equivalent is a common failure that visual clarity alone does not fix.