Violin

A box plot shows five numbers. Two groups can share the same median and quartiles and still look completely different: one tight and symmetric, the other bimodal or with a long tail. The violin shows that shape, and with a box, quartile lines or the points drawn inside it, the reader keeps the familiar summary as well.

Quartiles follow the definition you choose. The default is Excel / DAX inclusive, the same numbers PERCENTILE.INC returns in your model; Excel / DAX exclusive, Tukey's hinges and all nine Hyndman-Fan sample-quantile types are one click away. Whisker rules are 1.5 x IQR with an adjustable multiple, percentiles, mean plus or minus k standard deviations, or minimum to maximum.

The bandwidth comes from a named rule: Scott (the default), Silverman, or Silverman's rule of thumb. A smoothness control multiplies it when the shape needs to be calmer or sharper, densities can be trimmed to the data range so a violin never suggests values that were never observed, and violins can be scaled to equal area, equal width or the group's number of observations.

Because the rule is named, the chart can be stated in a methods section. A methods footnote on the visual says which quartile definition and bandwidth rule were used, and every statistic is golden-tested against independent reference implementations on nine fixture datasets. Degenerate inputs such as a single observation or zero spread return n/a instead of a misleading shape.

A raincloud puts a half violin, a slim box and the individual observations side by side, so a reader sees the shape, the summary and the raw data at once. It is the honest choice for small and medium samples, where a density alone can hide how few observations it is built on.

A ridgeline stacks one density per group along a shared value axis. It reads well with many groups, such as months, products or machines, and with mean, median and up to four percentile lines drawn per group it shows how the distribution drifts from one group to the next.

Premium answers the question every violin plot raises. Pairwise tests compare groups, or the two halves of a split violin, with Welch's t-test or the Mann-Whitney U test, adjusted across the pairs shown by Holm or Bonferroni; an overall test runs Welch's ANOVA, one-way ANOVA or Kruskal-Wallis. Results appear as brackets with stars or as letters, and too many brackets to read switch to letters automatically.

Premium also adds six kernels, a fixed bandwidth and bounded support by reflection for values with a natural floor or ceiling, confidence intervals, weights, spec limits typed or from measures with Cp, Cpk, Cpl, Cpu, Cpm, Pp and Ppk, and small multiples. Every premium feature previews free in edit mode; readers without a licence always see the free chart, never a blocked visual or a watermark.