When should I use a violin plot?

Comparing distribution shapes across groups ● Revealing bimodal distributions ● Showing skew or asymmetry that a box plot would hide

When should I avoid a violin plot?

Small samples (rule of thumb: n<20 per group makes the KDE unreliable) ● When audience is unfamiliar with the form

What are common mistakes with violin plots?

Wrong bandwidth (the KDE smoothing parameter — bin width is the histogram analogue) ● Ignoring outliers

What data does a violin plot need?

Minimum 10 rows, 1 columns (number). At least ~20 observations per group is recommended; below that the KDE shape is unreliable.

Distribution

Violin Plot

Q: What are common mistakes with violin plots?

Wrong bandwidth (the KDE smoothing parameter — bin width is the histogram analogue) ● Ignoring outliers

Q: What data does a violin plot need?

Minimum 10 rows, 1 columns (number). At least ~20 observations per group is recommended; below that the KDE shape is unreliable.

Combines box plot with kernel density to show full distribution shape.

Response Distribution by Treatment

Clinical trial results

View data (60 rows)

Chart data table: Response Distribution by Treatment
Group	Value
Control	42
Control	44
Control	46
Control	47
Control	48
Control	48
Control	49
Control	50
Control	50
Control	51
Control	52
Control	52
Control	53
Control	54
Control	55
Control	47
Control	49
Control	51
Control	53
Control	46
Treatment A	56
Treatment A	58
Treatment A	60
Treatment A	61
Treatment A	63
Treatment A	63
Treatment A	64
Treatment A	65
Treatment A	66
Treatment A	67
Treatment A	68
Treatment A	69
Treatment A	70
Treatment A	72
Treatment A	75
Treatment A	62
Treatment A	64
Treatment A	66
Treatment A	68
Treatment A	61
Treatment B	68
Treatment B	71
Treatment B	74
Treatment B	76
Treatment B	78
Treatment B	79
Treatment B	80
Treatment B	81
Treatment B	82
Treatment B	83
Treatment B	85
Treatment B	86
Treatment B	88
Treatment B	90
Treatment B	93
Treatment B	77
Treatment B	80
Treatment B	83
Treatment B	85
Treatment B	75

Make a violin plot with your data

Use a violin plot when…

Comparing distribution shapes across groups
Revealing bimodal distributions
Showing skew or asymmetry that a box plot would hide

Avoid when…

Small samples (rule of thumb: n<20 per group makes the KDE unreliable)
When audience is unfamiliar with the form

Data it needs

Property	Value
Min Rows	10
Min Columns	1
Column Types	number
Notes	At least ~20 observations per group is recommended; below that the KDE shape is unreliable.

Visual anatomy

Marks

area (mirrored)embedded box-plot

Channels

width (density)position-ymedian tick

Axes

x-categoricaly-quantitative

Guiding principles

PerceptionThe Encoding Hierarchy
Width encodes density, position encodes value

Consider instead

Box Plot

Simpler and more widely understood

Ridgeline Plot

For many groups stacked vertically

Common mistakes

Wrong bandwidth (the KDE smoothing parameter — bin width is the histogram analogue)
Ignoring outliers

History

Introduced by Hintze and Nelson in 1998 as an improvement over box plots.

Accessibility notes

Provide summary statistics as text alternative.

Got data? Let's see what works.

Drop your CSV. You'll get a Violin Plot plus four alternatives - ranked by which one actually fits your data best.

Try it free