Overview
ISO 16269-6:2014 - "Statistical interpretation of data - Part 6: Determination of statistical tolerance intervals" specifies procedures for constructing statistical tolerance intervals that, with a stated confidence level (1 − α), contain at least a specified proportion p of a population. The standard covers both one‑sided (upper or lower) and two‑sided tolerance intervals, provides parametric and distribution‑free methods, and includes procedures for multiple normal samples with a common unknown variance.
Key topics and requirements
- Parametric method (normal assumption): Procedures and forms for constructing tolerance intervals when the characteristic is assumed to follow a normal distribution. Uses specific forms (Forms A, B, C in Annex B) and provides k‑factors (Annex A, C, D) required to compute limits.
- Distribution‑free method: Nonparametric procedures (Form D, Annex E, G) for any continuous distribution when nothing is known about its form.
- One‑sided and two‑sided intervals: Methods for intervals with a single limit (upper or lower) and intervals with both limits; includes treatment for multiple samples with common unknown variance.
- Confidence and coverage: Clarifies the interpretation of confidence level (1 − α) and coverage proportion (p) - the interval is constructed to contain at least p of the population with confidence 1 − α.
- Statistical assumptions: Observations must be independent for procedures to be valid. Parametric results rely on normality; parametric methods for other distributions are not covered.
- Supporting material: Worked examples (Section 5), annexes with exact k‑factors, computation guidance (Annex F), and normative references (ISO 3534 series).
Applications and users
ISO 16269-6 is practical for:
- Quality engineers and process managers using statistical tolerance intervals to compare process capability with specification limits (upper U, lower L).
- Statisticians and data analysts designing tolerance intervals for reporting product conformity and uncertainty.
- Acceptance sampling by variables, where tolerance limits relate to acceptable quality limits (AQL).
- Regulatory bodies and testing laboratories that require documented methods to assert that a proportion of items meets requirements with a specified confidence.
- Reliability and product development teams estimating population coverage for performance characteristics.
Practical benefits include objective, reproducible methods for setting tolerance limits, selecting one‑sided vs two‑sided approaches, and choosing parametric vs distribution‑free procedures depending on data distribution.
Related standards
- ISO 3534-1:2006 and ISO 3534-2:2006 (vocabulary and symbols for statistics) - normative references used by ISO 16269-6.
- Other parts of ISO 16269 (e.g., Part 4 on outliers, Part 7 on medians, Part 8 on prediction intervals) provide complementary statistical guidance.
Keywords: ISO 16269-6, statistical tolerance intervals, parametric, distribution-free, confidence level, k‑factors, normal distribution, process capability, quality engineering.