Strip Plot

Key points

• 'Jitter' is often added to separate overlapping points, revealing densities.
• It is essentially a scatter plot where one axis is categorical.
• Strip plots are great for small to medium datasets where box plots might hide important details like sample size or underlying clusters.

Practical guidance

Showing every point is increasingly a journal expectation, not a stylistic choice - summary-only bar or box plots have been repeatedly shown to hide fabricated-looking patterns, unequal sample sizes, and outliers that drive the statistics, and a strip plot is the cheapest defense. The jitter is random noise added purely for legibility: keep it narrow enough that categories stay visually separate, and remember the horizontal position within a strip means nothing - readers sometimes over-interpret it. seaborn's stripplot(x='group', y='value', jitter=0. 2, alpha=0. 5) covers most cases; when you want jitter that reflects local density instead of randomness, swarmplot arranges points deterministically with no overlap, but it stops scaling around a thousand points per category, at which point return to a jittered strip with lower alpha or subsample. The strongest pattern in practice is the overlay: a strip plot on top of a de-emphasized box plot (draw the box first, light gray, showfliers=False so outliers aren't drawn twice) gives readers both the individual observations and the quartile summary in one panel. For paired or repeated measures, connect each subject's points across categories with thin lines - the within-subject trend that a plain strip plot destroys is often the entire finding.