ISO 11843-3:2003

Overview

ISO 11843-3:2003, "Capability of detection - Part 3", defines a statistical method to estimate the critical value of the response variable when no calibration data are available. It provides a practical procedure to decide whether a measured response from a test sample exceeds the distribution of responses attributable to a chosen reference (basic) state in which the net state variable is effectively zero. The method assumes data are normal or near-normal and is intended for cases where many blank/reference measurements can be made but measurements of true low-level actual states are scarce.

Key topics and requirements

• Scope and assumptions
  • Applicable when no calibration curve is used and the reference state can be prepared so that the net state variable ≈ 0.
  • Data are assumed to be normal or near-normal.
• Experimental design
  • Use the same complete measurement method for reference replicates and test samples.
  • Randomize actual-state measurements among basic-state replicates.
  • Do not discard or alter negative response values arising from measurements.
• Replication and uncertainty
  • Replicate the basic-state measurement J times to estimate the mean (yb) and standard deviation (sb). About 30 replicates is recommended to achieve a reasonably stable estimate of sb.
  • If J is small, use the chi-square distribution to assess uncertainty in sb.
• Computation of critical value
  • The critical value yc is computed from the basic-state mean and variance, using the appropriate quantile:
    • If σ0 known: use the normal quantile (z).
    • If σ0 estimated by sb: replace z with the Student’s t quantile with ν = J − 1 degrees of freedom.
  • The procedure allows for K replicate measurements on actual states; reporting should include J, K, sb, yb, chosen α (default 0.05) and the computed yc.
• Reporting
  • Explicitly state experimental parameters (J, K, α), basic-state mean and standard deviation, and the critical value.

Applications and users

• Who uses it: metrology and analytical laboratories, quality managers, method developers, calibration laboratories, and regulators.
• Typical applications:
  • Determining whether a low-level signal (e.g., trace analyte) is statistically distinguishable from blank/background when calibration at low levels is impractical.
  • Studies where preparing many blank/reference samples is feasible but obtaining many low-level spiked samples is difficult.
  • Establishing detection decision thresholds in chemical analysis, environmental monitoring, and quality control.

Related standards

• ISO 11843-1 (Terms and definitions)
• ISO 11843-2 (Methodology in the linear calibration case)
• ISO 5479 (Tests for departure from normality)
• ISO 5725-2 (Repeatability and reproducibility)
• ISO 11095 (Linear calibration using reference materials)
• ISO Guide 30; ISO 3534 (Statistics vocabulary)

Keywords: ISO 11843-3:2003, capability of detection, critical value, response variable, no calibration data, detection limit, Student’s t, replication, reference state, measurement uncertainty.