Find the right method for your analysis

Cross-cutting checks, whatever method you land on

• Independence. Clustered or repeated observations (many trips per person, many speeds per segment) break the independence most tests assume. Aggregate to one row per unit, or use mixed-effects models or cluster-robust standard errors.
• Effect size next to significance. Large transport samples make trivial differences significant. Report Cohen's d, eta-squared, or Cramer's V so the reader sees whether the difference matters.
• Multiple comparisons. A few planned contrasts: Holm or Bonferroni. Many exploratory ones: Benjamini-Hochberg (FDR).
• No post-hoc power. After a non-significant result, report the confidence interval and effect size, not observed power.