get_dblCalls_ROC: Derive doublet calls from classification probabilities

In kostkalab/scds: In-Silico Annotation of Doublets for Single Cell RNA Sequencing Data

Derive doublet calls from classification probabilities

Description

Given class probabilities (or scores) discriminating real data from artificial doublets, derive doublet calls. Based on selecting a ROC cutoff, see The Inconsistency of ‘‘Optimal’’ Cutpoints Obtained using Two Criteria basedon the Receiver Operating Characteristic Curve, (doi).

Usage

get_dblCalls_ROC(scrs_real, scrs_sim, rel_loss = 1)

Arguments

scrs_real	numeric vector, the scores for the real/original data
scrs_sim	numeric vector, the scores for the artificial doublets
rel_loss	numeric scalar, relative weight of a false positive classification compared with a false negative. Default:1 (same loss for fp and fn).

Value

numeric, vector containing the (estimated) number of doublets, the score threshold and the fraction of artificial doublets missed (false negative rate, of sorts)

kostkalab/scds documentation built on Oct. 4, 2022, 6:56 p.m.