ISO 5479-1997 PDF

Full title and description

St ISO 5479-1997 — Statistical interpretation of data — Tests for departure from the normal distribution. This international standard specifies a set of graphical and numerical procedures to assess whether a sample of observations departs from a normal (Gaussian) distribution. It provides guidance on selection and application of tests, interpretation of results and practical recommendations for sample sizes, transformations and presentation.

Abstract

ISO 5479:1997 gives methods for testing the null hypothesis that a sample of independent observations comes from a single normal distribution. The document covers graphical methods (normal probability plots), moment-based tests (skewness, kurtosis), regression and characteristic-function tests, recommendations on directional vs. omnibus testing, required significance levels and practical computation rules. It is intended to help analysts decide whether data may be treated as approximately normal or whether transformation or alternative models are needed.

General information

• Status: Published — confirmed current (last systematic confirmation carried out in 2021).
• Publication date: May 1997 (1997-05).
• Publisher: International Organization for Standardization (ISO).
• ICS / categories: 17.020; 03.120.30.
• Edition / version: Edition 1 (1997).
• Number of pages: 33.

Scope

The standard applies to samples of independent observations (recommended sample size eight or more) and gives procedures to test for departure from normality for either the raw observations or simple functions of them (for example logarithms or square roots). It treats graphical methods, moment tests, regression tests and characteristic-function tests, and excludes grouped-data chi-squared tests (because grouping causes loss of information). The standard also specifies significance levels, required numerical precision during calculations and guidance about transformations or splitting into subpopulations when appropriate.

Key topics and requirements

• Normal probability plot (graphical method) and interpretation guidance, including an annex with blank normal probability graph paper.
• Moment-based tests (measures of skewness and kurtosis) and their use as directional tests.
• Regression-type and characteristic-function tests (omnibus and directional testing strategies).
• Recommendation that tests are generally ineffective for very small samples — restricted to n ≥ 8.
• Guidance on choosing directional vs. omnibus tests depending on prior expectations about the form of departure.
• Specification of typical significance levels (α = 0.05 and α = 0.01) and tables of critical values for common tests.
• Numerical requirements: use at least six significant digits for intermediate calculations; avoid premature rounding.
• Advice on data transformation and on identifying and handling mixtures of subpopulations.

Typical use and users

This standard is used by statisticians, quality engineers, process-control and reliability analysts, researchers and laboratory personnel who need to verify the normality assumption before applying parametric methods (for example control charts, t-tests, ANOVA, capability studies). It is also useful in industrial quality improvement (Six Sigma, process capability analysis), metrology and academic research where objective, standardized normality assessment is required.

Related standards

ISO 5479 is part of the ISO statistical interpretation family and cross-references ISO statistical vocabulary and symbol standards (for example ISO 3534-1). It is often published alongside or implemented with national/adopted equivalents (for example regional or national adoptions and identical publications in other standards bodies) and is related conceptually to standards covering process capability, control charts and hypothesis testing.

Keywords

normality test, normal probability plot, skewness, kurtosis, moment tests, regression tests, characteristic function tests, omnibus test, directional test, sample size, statistical interpretation, hypothesis testing.

FAQ

Q: What is this standard?

A: ISO 5479:1997 is an international standard that specifies graphical and statistical methods to test whether a sample departs from the normal distribution.

Q: What does it cover?

A: It covers normal probability plots (graphical checks), moment-based tests (skewness and kurtosis), regression and characteristic-function tests, recommendations on sample size, significance levels (commonly 0.05 and 0.01), computation precision and guidance on transformations or treating mixed subpopulations.

Q: Who typically uses it?

A: Statisticians, quality and process engineers, laboratory analysts, researchers and anyone who must validate the assumption of normality before applying parametric statistical methods.

Q: Is it current or superseded?

A: The edition was published in May 1997 (Edition 1). The publication record shows the text was subject to systematic review and was confirmed as the current version in 2021; users should check their national standards body for the most recent confirmation or any national amendments before citation in regulatory contexts.

Q: Is it part of a series?

A: ISO 5479 is part of ISO’s statistical interpretation guidance and is intended to be used alongside other ISO statistical standards (for example ISO vocabulary standards for statistics and related process-capability and control-chart standards).

Q: What are the key keywords?

A: Normality test; normal probability plot; skewness; kurtosis; moment tests; omnibus test; directional test; sample size; statistical interpretation.