Overview
ISO 24185:2022 - Evaluation of the uncertainty of measurements from a stationary autocorrelated process provides a standardized method to evaluate the standard uncertainty of a process mean when successive measurements are possibly autocorrelated but form a stationary time series. The standard explains how to estimate the sample mean, autocovariance and autocorrelation, gives practical rules and statistical tests to validate the stationarity/autocorrelation assumptions, and shows how to relate that uncertainty to other Type A and Type B uncertainty components.
Key topics and requirements
- Stationary processes: Defines weak (covariance) stationarity - constant mean and variance, and autocovariance depending only on lag τ.
- Autocovariance and autocorrelation (γ(τ), ρ(τ)): Procedures to estimate γ(τ) and the sample autocorrelation ρ̂(τ) from discrete, equally spaced data.
- Uncertainty of the sample mean: Shows how autocorrelation affects the standard uncertainty of the mean and why the usual s/√N formula can be inappropriate when data are correlated.
- Tests and practical rules:
- Use the sample autocorrelation function (ACF) with confidence bands (approx. ±1.96/√N) to screen for white noise vs autocorrelation.
- Practical guidance such as recommending N ≥ 50 and useful lag range (|τ| ≤ N/4) for reliable ACF estimation.
- Approximate standard deviations for sample autocorrelations: σ̂ ≈ 1/√N for lag 1; extended formulae that include sums of earlier autocorrelations for higher lags.
- Additional cases covered:
- Incorporation of Type B uncertainty contributions.
- Treatment of weighted means.
- Includes tests for validity of the stationary/autocorrelation assumptions.
- Illustrative material: Annex A provides three practical examples to demonstrate application.
Practical applications and users
ISO 24185:2022 is directly applicable where repeated measurements are collected over time and positive autocorrelation is likely:
- Calibration and metrology laboratories evaluating mean measurement uncertainty.
- Process industries with continuous production (chemical, food, manufacturing) monitoring quality characteristics.
- Biomedical and environmental monitoring where biological or temporal effects induce autocorrelation.
- Quality engineers, statisticians, data scientists and regulatory bodies implementing measurement uncertainty assessments and statistical process control.
Benefits:
- More accurate uncertainty estimates when measurements are autocorrelated.
- Better decision-making for conformity assessment, control limits, and reporting of measurement results.
Related standards
- ISO 3534-2 (Statistics - Vocabulary and symbols - Part 2)
- ISO/IEC Guide 98-3:2008 (GUM) - guidance on measurement uncertainty
- ISO 7870-9 - statistical process control methods referenced for monitoring sequences
Keywords: ISO 24185:2022, measurement uncertainty, stationary autocorrelated process, autocorrelation, autocovariance, sample mean, Type A Type B uncertainty, time series, statistical process control.