Test-selection decision guide

Learning objectives

• Route a garment question to the right statistical test.
• Match test choice to data type, design and assumptions.
• Document H0/H1 and fallback plan before running the test.

Core idea

Test selection follows question, response type, groups, independence/pairing and assumptions — not the order of software menus.

Selection guide

• One mean vs target → 1-sample t.
• Two independent means → 2-sample t (Welch by default).
• Paired data (before/after same garment) → paired t.
• >2 means → one-way ANOVA (Kruskal-Wallis if assumptions fail).
• One/two proportions → z or chi-square.
• Association of categories → chi-square of independence.
• Variances → F or Levene.
• Continuous X↔Y → correlation / regression.

Garment-factory example

Comparing seam strength on the same garment before and after a needle change is paired data, not two independent samples; using a 2-sample t would ignore the pairing and inflate variance.

Method

1. Write parameter and H0/H1 in words.
2. Identify data scale and design.
3. Inspect plot and assumptions before selecting the test.
4. Document a non-parametric fallback.

Common mistakes

• Treating paired data as independent.
• Dichotomising continuous data prematurely.
• Running multiple unplanned tests without adjustment.