calibration_fusion: Fuses multiple LR systems and provides single calibrated...

In justinjhlo/fvclrr: Likelihood Ratio Calculation and Testing in Forensic Voice Comparison

Fuses multiple LR systems and provides single calibrated output

Description

Combines LR scores from (multiple) test systems to provide a single set of fused LR scores, based on a logistric-regression model trained with the training data input. Self-calibration based on training with the same set of data is possible, in which case the training data argument should be omitted. An R implementation of calibration_fusion.m by Hughes (2015) and train_llr_fusion_robust.m by Morrison (2009) (http://geoff-morrison.net/#TrainFus).

Usage

calibration_fusion(test_list, dev_list = NULL, log = TRUE, ...)

Arguments

test_list	A list of LR matrices to be calibrated and fused. The input must be a LIST. Any data frame or matrix will work, as long as it follows the same format as the output in MVKD_loop (a row for each suspect, a column for each offender, LR value not transformed in any way). The LR matrices should all be of the same dimensions.
dev_list	(Optional) A list of LR matrices to be trained on. The list must be of the same length as test_list and matrices within must correspond with those in test_list. If left blank, training will be based on test_list.
log	Boolean. LRs already log-transformed (TRUE; default) or not (FALSE)?

Value

A named list following the exact format as MVKD_loop: likelihood_ratio_matrix, cllr and eer.

justinjhlo/fvclrr documentation built on June 27, 2022, 11:19 a.m.