fragility_report: Per-case fragility report
In mispitools: Missing Person Identification Tools

fragility_report

R Documentation

Per-case fragility report

Description

Produces the case-level fragility summary described in §5 of Marsico & Egeland (in preparation): total weight of evidence, inclusion concentration index C_W^+, worst-case leave-one-out loss, residual support after removal of the most impactful marker, and an optional comparison against a pedigree-specific cutoff (typically produced by calibrate_concentration_cutoff).

Usage

fragility_report(per_marker_lrs, cutoff = NULL, probs = NULL)

Arguments

`per_marker_lrs`	Named numeric vector of per-marker likelihood ratios (strictly positive). Names become the marker labels in the report.
`cutoff`	Optional numeric scalar. If supplied, the report flags cases whose observed `C_W^+` exceeds `cutoff` as requiring a leave-one-out review before final reporting.
`probs`	Optional numeric in `(0, 1)`. The quantile level at which `cutoff` was calibrated. Used only to enrich the reportable sentence (`"below the 90th-percentile cutoff"` etc.); ignored if `cutoff` is `NULL`.

Details

The report is a direct operationalization of Paper 1 §5:

W is the combined weight of evidence in bans.
C_W^+ answers the question “what fraction of the inclusion support comes from a single marker?”
(1 - C_W^+) W is the residual support after the single-marker worst case, i.e.\ what the analyst would be left with if a successful challenge to the top marker were sustained.
The flag is raised when C_W^+ exceeds the pedigree-specific cutoff, which in turn is typically the 90th percentile of the H_p simulation distribution produced by calibrate_concentration_cutoff.

Value

A list with elements

total_log10_lr: Total weight of evidence W = \sum_k \log_{10}\mathrm{LR}_k (in bans).
concentration: Inclusion concentration index C_W^+ \in [0, 1].
top_marker: Name of the single most impactful supporting marker.
top_log10_lr: Log-LR of the top marker (in bans).
residual_log10_lr: Total support that survives worst-case removal of the top marker, (1 - C_W^+) W.
flag: Logical. TRUE if cutoff was supplied and C_W^+ > \text{cutoff}; NA otherwise.
statement: Character scalar. A ready-to-paste natural-language fragility statement for the case report.

References

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

Examples

# Balanced case: 15 markers, similar contributions
set.seed(1)
lrs <- setNames(runif(15, 1.5, 3.5), paste0("M", 1:15))
fragility_report(lrs)

# Concentrated case: one dominant marker
lrs2 <- setNames(c(1e5, rep(1.05, 14)), paste0("M", 1:15))
fragility_report(lrs2, cutoff = 0.20, probs = 0.90)

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