Distance between a derived cut and its FMO-anchored equivalent

The diagnostic the feature exists for. Positive means the sample's own density put the cut above the control; negative means below.

Usage

fmo_agreement(
  thr_all,
  fmo_thresholds,
  unc = NULL,
  agree_at = 1,
  disagree_at = 3
)

Arguments

thr_all

the thresholds table, carrying sample_id, marker, threshold and source

fmo_thresholds

data.frame of sample_id, marker, fmo_threshold, fmo_sample, and optionally fmo_verdict from control_cut(). A row whose fmo_threshold is NA is a control that was refused, and is reported as such rather than compared.

unc

optional thresholds element of run_gate_uncertainty(), used to scale the distance. A cut the control supplied has no uncertainty of its own, because the events it came from are in a separate tube, so those rows carry distance but not distance_in_u. The scaled verdict is available for the rows where the sample's own cut was the one applied.

agree_at

distance in uncertainties within which the two agree

disagree_at

distance beyond which they are reported as disagreeing

Value

a data.frame, or NULL

Details

HOW TO READ distance_in_u. It is the gap expressed in units of that threshold's own standard uncertainty, from threshold_uncertainty.csv. Within about one, the two methods agree to the precision either can claim, and the derived cut is corroborated by an independent experiment. Beyond about three, they disagree by more than either can explain and one of them is wrong: a derived cut far above the FMO is discarding real signal, and one far below is calling spillover positive.