Stacked violin plot — violin • plotomics

One row per feature, one violin per group. A box plot hides bimodality, which in single-cell data is usually the whole story: a gene expressed in half a cluster and silent in the other half has the same median as one expressed weakly everywhere. The violin shows the shape, and stacking rows on a shared x lets a marker panel be read down the page.

Usage

violin(
  data,
  grid,
  density,
  grids = NULL,
  median = NULL,
  features = NULL,
  groups = NULL,
  group_colors = NULL,
  violin_width = 0.85,
  scale_per_violin = FALSE,
  show_median = TRUE,
  show_feature_labels = TRUE,
  theme = NULL,
  width = NULL,
  height = NULL,
  element_id = NULL
)

Arguments

data: A data frame with feature and group key columns, one row per violin, in the order to draw them.
grid: Numeric vector, the shared evaluation grid, ascending.
density: Numeric matrix, violins x grid, of density values.
grids: Optional numeric matrix, features x grid, giving each feature its own y range. Without it every row shares grid, which lets one highly expressed feature compress the rest into flat lines.
median: Optional numeric vector, one median per violin, drawn as a tick.
features, groups: Character vectors fixing the row and column order. Factor feature / group columns supply them from their levels.
group_colors: One hex colour per group. NULL uses the categorical palette.
violin_width: Fraction of a cell's width the widest violin fills.
scale_per_violin: Scale each violin to its own maximum rather than the row's. Per-row is the default so groups stay comparable within a feature.
show_median, show_feature_labels: Toggle the median tick and row labels.
theme: Optional named list of theme overrides.
width, height: Widget dimensions (any valid CSS size).
element_id: Optional explicit DOM id.

Value

An htmlwidget object.

Details

The widget draws densities, it does not estimate them. Each violin arrives as a vector of density values on a shared grid, because kernel bandwidth choice changes what the figure claims and belongs with the data. violin_density() computes them from raw values with stats::density().

Examples

d <- violin_density(list(`CD3D|T` = rnorm(50, 2), `CD3D|B` = rnorm(50, 0)))
violin(d$data, grid = d$grid, density = d$density, median = d$median)