Skip to contents

Get Scatter Occupancy

Source: R/get_scatter_occupancy.R
get.scatter.occupancy.Rd

Determines the region of 2D scatter space (typically FSC vs SSC) that a sample's events actually occupy, using a density-relative threshold rather than a fixed instrument range. A 2D kernel density estimate is built over a robust, quantile-trimmed span of the data, so that a handful of extreme events cannot inflate the estimation grid and coarsen its resolution over the region that actually matters. Events falling in density bins below density.threshold of the peak density are treated as sparse noise or outlier events and excluded; the remaining "occupied" events define a data-driven working range for gating and plotting.

This identifies dense clusters generally, and does not by itself distinguish a real debris population from the population of interest – a distinct, well-populated debris cloud is typically just as dense as real cells or beads and will not be excluded here. Separating debris from the target population remains the job of the downstream target-maximum search (e.g. the bottom-left exclusion logic in do.gate()); this function only keeps sparse, spread-out outlier events from distorting the working scale used to set that search up.

Usage

get.scatter.occupancy(
  scatter.data,
  density.threshold = 0.02,
  grid.n = 128,
  max.events = 1e+05,
  trim.quantile = 0.001,
  bird.seed = NULL
)

Arguments

scatter.data

A matrix or data frame with (at least) two columns: Forward Scatter and Side Scatter, in that order.

density.threshold

Numeric 0-1, default 0.02. Bins with a kernel density below this fraction of the peak density are excluded as sparse.

grid.n

Numeric, default 128. Binning grid for the kernel density estimation.

max.events

Numeric, default 1e5. Maximum number of events used to fit the kernel density surface. Above this, a random subsample is used to fit the surface; every event is still evaluated against it.

trim.quantile

Numeric 0-0.5, default 0.001. Quantile trimmed from each end of the data range before setting the density grid limits, so a small number of extreme events cannot stretch the grid.

bird.seed

Integer, seed for reproducible subsampling. Default NULL skips seeding.

Value

A list with:

  • keep: Logical vector, length nrow(scatter.data), TRUE for events in an occupied (dense) region.

  • x.range, y.range: Two-element numeric vectors giving the occupied range on each axis, taken from the retained events.

  • density: The x/y/z kernel density surface used to determine occupancy (same structure as MASS::kde2d() output), useful for diagnostic plotting.

See also