Overview - ISO/TR 20693:2019 (Statistical methods for Six Sigma)
ISO/TR 20693:2019 provides guidance for distribution identification within Six Sigma implementations. The technical report documents graphical and numerical procedures to identify one‑dimensional, unimodal probability distributions - either continuous or discrete. It emphasizes exploratory data analysis (EDA), hypothesis testing (including goodness‑of‑fit and normality tests), and pragmatic decision steps that fit within the DMAIC lifecycle (recommended to start in the Measure phase and continue through Improve and Control).
Key topics and technical highlights
- Scope: One‑dimensional, single‑mode distributions; continuous or discrete data only.
- Exploratory Data Analysis (EDA): Use of descriptive statistics (mean, median, skewness, kurtosis, quartiles) and visual tools such as histograms, boxplots, stem‑and‑leaf, and Q‑Q plots to form distribution hypotheses.
- Distribution selection: Guidance to narrow candidate families based on process knowledge (e.g., positive‑only, discrete counts) and Occam’s Razor (favor simpler models like exponential, normal, Poisson when appropriate).
- Graphical and numerical procedures: Illustrated workflows for discrete and continuous cases, combining visual diagnostics with formal tests and goodness‑of‑fit methods.
- Practical workflow: Steps include stating objectives, formulating model theory, collecting/preparing/exploring data, selecting candidate distributions, performing goodness‑of‑fit testing, and drawing conclusions.
- Illustrative examples: Annexed case studies covering lottery uniformity tests, post‑release technical issues, software effort estimation, and warranty period determination to demonstrate applied techniques.
Applications - practical value
- Supports Six Sigma project teams in selecting appropriate parametric methods by verifying underlying distributional assumptions.
- Enables quality engineers, reliability engineers, data analysts, and statisticians to model process behavior, estimate life or failure distributions, and choose correct statistical tests for inference and control.
- Useful for tasks like process baseline characterization (Measure), root‑cause analysis (Analyse), validating improvements (Improve), and ongoing monitoring (Control).
- Helps avoid misleading automation in software packages by combining contextual process knowledge with formal statistical methods.
Who should use this standard
- Six Sigma practitioners (Green/Black Belts), quality and process improvement professionals, applied statisticians, reliability engineers, and data scientists involved in industrial or service‑domain process modelling and control.
Related standards
- ISO 3534‑1 - Statistics: vocabulary and symbols (referenced normative).
- ISO 16269‑4 - Statistical interpretation of data (boxplot terminology referenced).
Keywords: ISO/TR 20693:2019, Six Sigma, distribution identification, exploratory data analysis, goodness of fit, Q‑Q plot, histogram, normality test, DMAIC.