Interpreting Chi-Square Output in Stata: Independence, Goodness-of-Fit & Caveats
Chi-Square in Stata: Interpretation Guide
1. Test of Independence (tabulate, chi2)
- H₀: Variables independent. H₁: Dependent.
- Pearson χ² & p-value: p < 0.05 → reject H₀ (evidence of association). p ≥ 0.05 → fail to reject.
- Degrees of freedom (df): (rows-1)(cols-1); used for critical value.
- Expected vs. observed counts: Inspect for driving cells.
2. Goodness-of-Fit (csgof / chitest)
- H₀: Data follow specified distribution. H₁: Do not.
- Same p-value logic. Must declare expected distribution (uniform, Poisson, etc.).
3. Other Uses
- Proportion equality, variance tests—same principles.
Caveats
- Large N → trivial effects become significant.
- Expected counts <5 → use Fisher's exact (tabulate, exact).
- p-value ≠ effect size; report Cramér's V or phi for strength.
Example
Gender | Agree Disagree | Total
-------+--------------+------
Male | 120 80 | 200
Female | 150 100 | 250
-------+--------------+------
Total | 270 180 | 450
Pearson chi2(1) = 10.00 P = 0.0016
p = 0.0016 < 0.05 → significant association between gender and opinion.