Choose between independent, paired, or one‑sample t‑tests; use Welch when variances or sample sizes differ; and always report effect sizes with confidence intervals.

Designs & When to Use

  • Independent samples: two unrelated groups
  • Paired: repeated measures or matched pairs
  • One‑sample: compare mean to a reference value

Assumptions

  • Approximately normal residuals (robust for n ≥ 30/group)
  • Independent observations by design
  • Equal variances for pooled t‑test; if not, use Welch

Effect Sizes

  • Cohen’s d or Hedges’ g with 95% CIs
  • Report alongside p‑values to communicate practical magnitude

Alternatives

  • Mann‑Whitney U (independent) or Wilcoxon signed‑rank (paired) for strong non‑normality
  • Bootstrap CIs when assumptions are doubtful
Run t‑Test Back to Service Page