Violin Plot

A violin plot displays the full probability density of the data at each value, mirrored to create a symmetric shape. Unlike a box plot, it reveals bimodality, skewness, and gaps in the data. It is the preferred alternative to bar charts in modern biology publications and is recommended over box plots when n >= 20 and the distribution shape is scientifically meaningful.

STEP 1 —When to Use

Use a violin plot when: n >= 20 per group (KDE is reliable at this n); the distribution shape may be non-normal, bimodal, or skewed; you are submitting to a journal that has moved away from bar charts (Nature, eLife, PLoS Biology); you want to show both the summary statistic and the full distribution.

Do not use a violin plot when: n < 10 (KDE will produce an artificially smooth curve that misrepresents the data —use strip/jitter plot instead); you have very many groups (>8) where individual violins become too narrow; the bandwidth parameter choice is not disclosed (the shape can be misleading if bandwidth is too large or too small).

[Image placeholder: Three panels showing the same dataset (n=25, bimodal distribution) displayed as: (1) bar chart (mean ﾂｱ SEM —completely hides bimodality), (2) box plot (median and IQR —partially hides bimodality), (3) violin plot (KDE reveals both modes clearly). Each panel has identical group means.]

Bar chart and box plot hide bimodality (left, center). Violin plot reveals the complete distribution structure (right).

STEP 2 —Journal Requirement

Bandwidth parameter must be stated in the legend or methods: "Kernel density estimation with bandwidth selected by Silverman's rule of thumb."
Inner summary statistic (median, mean, or quartiles shown within the violin) must be labeled.
n per group and the statistical test result must accompany the figure.
eLife, Nature Methods, and Cell require a note if individual data points are not shown when n < 50.

STEP 3 —Licklider Implementation

Input

Continuous outcome and group variable
Bandwidth selection: Silverman (default), Scott, or user-specified
Inner display: box (median + IQR), points (individual data), quartile lines, or none

Output

KDE curve mirrored symmetrically per group
Inner box optional: median line + IQR box drawn inside violin
Individual data points overlaid (jitter) for n < 50 by default
Bandwidth parameter recorded in legend automatically
Scale: area-normalized (all violins same area) or width-normalized (all violins same max width)
Significance annotations positioned above violins

Style Options

Half-violin (raincloud plot: half violin + strip plot + box, stacked horizontally)
Split violin (two groups sharing one axis position —for paired comparisons)
Fill transparency for overlapping violins

[Image placeholder: Licklider violin plot with 3 groups (n=35 each). Shows area-normalized violins with inner box (median line + IQR), individual data points overlaid with jitter, n labeled, significance brackets, and bandwidth disclosed in the legend.]

Licklider violin plot: area-normalized violins, individual data overlay, inner box, significance brackets, and auto-disclosed bandwidth.

STEP 4 —Draft Output (Draft / Needs review)

Data are presented as violin plots showing the kernel density estimate
(bandwidth selected by Silverman's rule of thumb, area-normalized).
The median and interquartile range are shown within each violin.
Individual data points are overlaid (n = 35 per group). One-way ANOVA
with Holm post-hoc correction; overall F(2, 102) = 8.3, p = 0.0004;
pairwise results shown in figure.