Choose a threshold by trying every method and scoring the cuts alike

Usage

best_threshold(
  x,
  ladder = c(4, 6, 8, 12, 16, 24, 32, 48, 64),
  bins = 400L,
  min_quality = 0.05,
  fallback_q = 0.9,
  floor = -Inf,
  prefer = c("deepest", "lowest"),
  tail = c("upper", "lower")
)

Arguments

x: transformed values for one marker within the parent gate.
ladder: smoothing windows for the mindensity sweep, widest last.
bins: histogram resolution for the mindensity sweep. Default 400L.
min_quality: a candidate must sit in a trough at least this deep to be preferred over the tail rule. Default 0.05.
fallback_q: quantile used only when nothing else is usable.
floor: no candidate at or below this value is considered; see density_valley(). Default -Inf.
prefer: "deepest" (default) scores every candidate by the trough it sits in. "lowest", for a gate that removes only the negative mode, takes the lowest density valley that sits in a real trough, and falls back to the scored choice when no valley does; see density_valley().
tail: "upper" (default, also meaning nothing was declared) offers a tail gate on either side of the dominant mode and lets the trough decide; "lower", the analyst's declaration that the dominant mode is positive (thresholds: <marker>: {tail: lower}), offers only the lower one and takes it as the last resort too, as the default path does. Without this the declaration was ignored under --adaptive-gates.

Value

list(threshold, source, quality), where source names the method that won so a reader can see how each cut was placed.