The F distribution is a cornerstone of statistics tutoring, especially in ANOVA, regression, and variance comparison tests. Students often memorize the formula without understanding why the ratio of two scaled chi-square variables produces an F distribution. This page explains the construction, intuition, and exam‑relevant logic behind the F distribution.
If \(U \sim \chi^2_{d_1}\) and \(V \sim \chi^2_{d_2}\) independently, then \[ F = \frac{U/d_1}{V/d_2} \sim F(d_1, d_2). \]
Why does this ratio produce an F distribution? Because each chi-square variable is itself a sum of squared standard normals. Dividing by degrees of freedom converts each into an unbiased variance estimator. The ratio of two independent variance estimators follows an F distribution. This structure explains why F tests compare variances across groups or models.
- Start with two independent chi-square variables. \(U \sim \chi^2_{d_1}\), \(V \sim \chi^2_{d_2}\).
- Divide each by its degrees of freedom. \(U/d_1\) and \(V/d_2\) are variance estimators.
- Form the ratio. \(F = (U/d_1) / (V/d_2)\).
- Use independence. Independence ensures the ratio has a clean distributional form.
- Apply the gamma connection. Chi-square variables are gamma; the ratio of scaled gammas yields an F distribution.
- Interpret the result. The F statistic compares two sources of variation.
Suppose:
- \(U = 12.4\) with \(d_1 = 4\)
- \(V = 18.0\) with \(d_2 = 6\)
Then:
\[ F = \frac{12.4/4}{18.0/6} = \frac{3.1}{3.0} \approx 1.033. \]
This value would be compared to an \(F(4,6)\) distribution in hypothesis testing.
- Thinking the F distribution is symmetric (it is right‑skewed).
- Forgetting that independence of the chi-square variables is required.
- Confusing F tests with t tests (t² = F with 1 numerator df).
- Misinterpreting the F statistic as a variance itself