calibrate_concentration_cutoff: Calibrate pedigree-specific concentration cutoff

In mispitools: Missing Person Identification Tools

Calibrate pedigree-specific concentration cutoff

Description

Simulates the distribution of the inclusion-fragility concentration index C_W^+ under the prosecution hypothesis H_p for a given pedigree and marker panel, and returns the chosen upper quantile as a case-flagging cutoff. Profiles whose observed C_W^+ exceeds the cutoff should trigger a leave-one-out review before reporting.

Usage

calibrate_concentration_cutoff(
  reference,
  missing,
  numsims = 1500,
  probs = 0.9,
  seed = 123,
  numCores = 1
)

Arguments

reference	A pedtools::ped object with markers attached and founder profiles simulated, as expected by sim_lr_genetic.
missing	Character or numeric. ID of the missing person in the pedigree.
numsims	Integer. Number of H_p simulations used to build the empirical C_W^+ distribution. Default 1500.
probs	Numeric in (0, 1). Quantile of the H_p distribution to use as the cutoff. Default 0.90 (matches the paper convention).
seed	Integer. Random seed passed through to sim_lr_genetic.
numCores	Integer. Cores for the underlying forrel::profileSim call.

Details

The calibration uses only the H_p branch of sim_lr_genetic: for each simulated matching profile, the per-marker LR vector is extracted and passed to concentration_index_positive. Simulations with no positive per-marker support (empty numerator) are dropped before taking the quantile.

The default probs = 0.90 matches the cutoff convention used in Marsico & Egeland (in preparation): a 10\ H_p is deemed acceptable in exchange for catching concentrated cases where the combined LR depends heavily on a single marker.

Value

A list with elements

cutoff

The probs-quantile of C_W^+ under H_p.

probs

The quantile level used.

distribution

Numeric vector of simulated C_W^+ values, length numsims.

total_log10_lr

Numeric vector of simulated \log_{10}\mathrm{LR} totals under H_p.

numsims

The number of simulations actually used (profiles with non-positive total support are discarded).

References

Marsico, F. L. & Egeland, T. (in preparation). Belief dynamics during the investigative process.

See Also

sim_lr_genetic, concentration_index_positive, fragility_report.

Examples

## Not run: 
  library(pedtools); library(forrel)
  x <- linearPed(2)
  x <- setMarkers(x, locusAttributes = NorwegianFrequencies[1:15])
  x <- profileSim(x, N = 1, ids = 2)
  cal <- calibrate_concentration_cutoff(x, missing = 5,
                                        numsims = 500, probs = 0.90)
  cal$cutoff

## End(Not run)

mispitools documentation built on Aug. 26, 2026, 1:08 a.m.