Minimum Detectable Presence Table¶
Artifact — instantiates Null Finding Warrant Calibration
States the smallest detectable target level, effect size, defect rate, incidence, or trace intensity.
A null finding has a floor, and the floor is the whole story. The Minimum Detectable Presence Table is the lookup artifact that states, for the method that was used, the smallest target level it could reliably have caught — and therefore the ceiling below which the target may still be hiding, unseen. Its defining move is to convert "we found nothing" into a bounded quantity: nothing above X. It answers the one question that makes a null usable — "how small could the target be and still have escaped us?" — by reading the detection-power curve at a stated confidence and tabulating the answer across whatever axis matters. Unlike the likelihood ratio, which prices how surprising the null is against a prior, the table states a level — a detection floor and an upper bound that hold independently of anyone's prior belief.
Example¶
A radio survey of a nearby star sweeps a band of frequencies for a narrowband artificial signal and detects nothing. "No signal" is nearly content-free. The table gives it content. From the survey's sensitivity — antenna gain, integration time, bandwidth — it reads off the faintest transmitter the search could have caught at ninety percent confidence, and tabulates it across frequency and assumed distance. The null becomes a row of ceilings: no narrowband transmitter above a given effective radiated power, within this range, across these frequencies, at ninety percent detection. That is a genuine, publishable result — an upper bound on transmitter power — rather than a shrug. It also makes the limits of the search honest: at the band edges, where sensitivity drops, the ceiling rises, so the same "nothing" is a much weaker statement there, and the table shows precisely where.
How it works¶
- Fix the confidence that defines "detectable." Choose the detection probability — say ninety or ninety-five percent — at which a level counts as reliably catchable.
- Read the detection-power curve. At that confidence, read the smallest target level the method detects; this is the minimum detectable presence.
- Tabulate across the relevant axis. Repeat across frequency, mass, distance, dose, or whatever axis the target varies along, so the floor is stated where it actually differs.
- Recast the null as a ceiling. Express the finding as "target below the threshold at confidence c," an upper-bound statement rather than a bare absence.
Tuning parameters¶
- Confidence level — the detection probability that defines the floor. A higher bar yields a more conservative (higher) detectable level and a more defensible ceiling; a lower bar flatters the search.
- Axis granularity — how finely the table samples frequency, dose, or size. Finer rows show where sensitivity sags but cost more to compute and justify.
- Single limit vs. full curve — a headline number or the whole sensitivity curve. The curve is honest about where the search was weak; the single limit is easier to quote and easier to misapply.
- Units — the scale the threshold is stated in, chosen to match the decision (regulatory limit, biological effect size, actionable defect rate).
When it helps, and when it misleads¶
Its strength is that it turns a bare null into an honest, decision-ready ceiling: "nothing above X, at this confidence, under these conditions" is a statement a regulator, editor, or engineer can actually use, and it degrades gracefully because the conditions travel with the number.
Its failure mode is quoting the best-case limit as if it held everywhere — lifting the deepest detection limit from the sweet spot of the curve and applying it at the edges where sensitivity was far worse — or stating a limit with no confidence level attached, which makes the ceiling meaningless. The idea is the generalization of the analytical-chemistry limit of detection, the LOD,[n1] and inherits its rule: a detection limit is only defined together with the confidence and conditions it holds under. The guarding discipline is to publish the confidence and the conditions beside every threshold, and to prefer the curve over a single flattering number.
How it implements the components¶
The table realizes the how-small-could-it-be side of the archetype:
minimum_detectable_presence_threshold— the table's core cell: the smallest target level the method reliably detects, at the stated confidence.upper_bound_statement— it recasts the null as "target below the threshold," a bounded ceiling rather than a bare absence.detection_power_model— the table is the detection-power curve read at the actionable confidence level and tabulated across the relevant axis.
It states what the null could have caught but does not weigh how surprising the null is or issue the overall verdict: the Bayes-factor reasoning and the displayed absence_warrant_grade belong to its nearest twin, the Likelihood Ratio for Non-Detection — the table supplies the ceiling; the ratio and the Null Finding Warrant Memo turn it into a graded conclusion.
Related¶
- Instantiates: Null Finding Warrant Calibration — the table supplies the archetype's upper-bound form of a calibrated null.
- Sibling mechanisms: Null Finding Warrant Memo · Search Sensitivity Matrix · Detection Power Checklist · Negative Test Interpretation Protocol · Likelihood Ratio for Non-Detection · Coverage Map and Blind-Spot Review · Silent Monitor Assurance Review
Editorial Notes¶
Form Classification¶
Form family: Analysis, Modeling & Optimization
Rationale: The mechanism reads a detection-power curve at a fixed confidence and computes threshold values across the relevant axis, yielding an inferential upper bound on absence.
Nearest alternative: Representation, Specification & Plan — The results are published in a table, but the operative value comes from the threshold inference rather than static presentation.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Chemistry & Materials Science
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: A table of the smallest detectable concentration, trace, or presence descends from analytical chemistry's limit-of-detection practice. Statistics defines decision error and signal thresholds; engineering supplies instrument sensitivity and calibration.
Related originating lineages:
- Engineering & Design — Metrology and quality inspection operationalized detection thresholds for defects and traces.
- Statistics & Experimental Design — Retained as a formative lineage independently identified as primary: Detection limits and smallest reliably detectable rates derive from statistical hypothesis testing and sampling.
Review resolution: IUPAC defines detection limit as the smallest concentration or amount detectable with stated confidence. That exact concept anchors the mechanism in analytical chemistry, even though the table generalizes it to effects and incidence. The alternates are retained only as formative or independently established origins, not because the mechanism can be applied there. origin_mode=cross_disciplinary_synthesis states the provenance relationship; domain_reach=multi_domain separately records breadth because independent established uses occur in several fields. confidence=high reflects the strength and specificity of the evidence; encyclopedia_synthesis=false because the entry generalizes an established mechanism without inventing a new composite.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
- https://goldbook.iupac.org/terms/view/L03540 — IUPAC's authoritative Gold Book defines the limit of detection.
- https://www.epa.gov/choose-fish-and-shellfish-wisely/lab-quality-assurance-and-quality-control — EPA laboratory QA guidance documents method detection limits and analytical controls.
Notes¶
[n1] The limit of detection (LOD) in analytical chemistry is the lowest quantity of a substance an assay can distinguish from a blank at a stated confidence. It is meaningless without that confidence and the measurement conditions — which is exactly why the table quotes them beside every threshold rather than reporting a floating "nothing found." ↩