Module 4 · Lesson 8 of 22
Distribution shape and normality assessment
Learning objectives
- Recognise skew and multimodality from plots.
- Interpret normality tests without over-reaction.
- Choose transformations, distribution models or non-parametric methods.
Assumption, not requirement
Normality is an assumption of some tests, not a business goal. Large samples detect trivial deviations; small samples miss real ones.
Combine histogram, probability plot, subgroup mixing check and process knowledge before deciding.
Garment-factory example
Repair time is right-skewed; mixed sizes create multiple peaks in measurement data. A log transformation stabilises variance for cycle-time analysis.
Method
- Verify data cleanliness and mixture sources.
- Plot histogram and probability plot.
- Run a normality test as a supplement.
- Choose transformation, distribution model or non-parametric method.
Common mistakes
- Declaring normality solely because p > 0.05.
- Deleting genuine tails.
Knowledge check
Pick one answer per question. Explanations appear after you submit.
1. Normality is:
2. A very large sample often produces:
Author: Sanjeewa Dehiwalage · Last reviewed: 2026-07-21