What Is the F-Distribution? A Plain-English Guide

What Is the F-Distribution?

To build intuition before the math: imagine you draw two independent samples from a normal population and compute the variance of each. Divide the larger variance by the smaller one. The resulting ratio follows an F-distribution. If the two populations actually have the same variance, this ratio should hover around 1. The further it drifts above 1, the more evidence you have that something is different between the groups.

The distribution was named after Ronald A. Fisher, who introduced it in 1924 as the "variance ratio distribution." George W. Snedecor later popularized it and named it F in Fisher's honor. Today it is a standard tool taught in every introductory statistics course and used in software from R to Python to SPSS.

You can read more about the underlying framework in the hypothesis testing guide on Statistics Fundamentals.

The Variance Ratio: A Plain-English Analogy

Think of it this way. You want to know whether three different fertilizers produce the same average crop yield. You plant ten plots with each fertilizer and measure the yield. Now you have two sources of variation:

• Variation between groups — the differences in average yield across the three fertilizer types
• Variation within groups — the natural scatter in yield among plants that received the same fertilizer

If the fertilizers have no real effect, both sources of variation should be roughly equal, giving an F-ratio near 1. If one fertilizer genuinely outperforms the others, the between-group variation grows large relative to within-group noise, pushing the F-ratio well above 1. The F-distribution tells you exactly how large that ratio would have to be before you can confidently rule out chance.

Shape of the F-Distribution

The F-distribution is always right-skewed and bounded below by zero. Its exact shape depends entirely on two parameters: d₁ (numerator degrees of freedom) and d₂ (denominator degrees of freedom). As these numbers grow, the distribution becomes less skewed and begins to approach a normal distribution — but it never becomes symmetric in the way the t-distribution does.

Why Is the F-Distribution Always Positive?

Interactive Curve Explorer

How Degrees of Freedom Change the Shape

When d₁ = 1, the distribution is J-shaped (highest near zero, falling monotonically). When d₁ ≥ 2, a modal peak appears to the right of zero. As both degrees of freedom grow large, the mean approaches 1 and the variance approaches 0 — the distribution concentrates near 1 because, under the null hypothesis, you expect the variance ratio to equal 1.

Degrees of Freedom in the F-Distribution

Unlike the t-distribution (one parameter) or the chi-square distribution (one parameter), the F-distribution has two separate degrees of freedom. Getting them right is the first practical step in any F-test.

Degrees of Freedom in One-Way ANOVA

One-way ANOVA partitions the total variation in your data into two components. Each component has its own degrees of freedom, and their ratio produces the F-statistic.

Here k is the number of groups and n is the total number of observations across all groups. Notice that d₁ + d₂ = (k−1) + (n−k) = n−1, which equals the total degrees of freedom. This additive property is the mathematical reason the ANOVA partition works.

The F-Statistic Formula

The F-statistic is the test statistic you compute from your data and then compare against the F-distribution to get a p-value.

Breaking Down the Components

SS_between (sum of squares between groups) measures how much each group mean deviates from the grand mean, weighted by group size:

SS_within (sum of squares within groups) measures natural variability inside each group — the variation you cannot attribute to group differences:

The Decision Logic

How to Read an F-Distribution Table

An F-table gives you the critical value F*(d₁, d₂, α) — the cutoff above which you reject H₀. Most tables are organized for a specific α (commonly 0.05 or 0.01), with d₁ across the columns and d₂ down the rows.

Sample F-Table (α = 0.05, Right-Tail)

Three-Step Table-Reading Process

Worked Examples

Three examples follow, each covering a different real-world scenario where the F-distribution applies. The calculation steps are written out fully so you can follow along with your own data.

Example 1: One-Way ANOVA (Three Teaching Methods)

Example 2: Testing Equality of Variances

Example 3: Global F-Test in Multiple Regression

Assumptions of the F-Test and ANOVA

The F-test is valid only when a set of conditions holds. Violating these can make your p-value unreliable — either too small (false positives) or too large (false negatives).

F-Distribution vs. t and Chi-Square Distributions

The F-distribution is part of a connected family of distributions. Understanding these relationships deepens your intuition for when each test applies.

The Mathematical Connection to Chi-Square

If X₁ ~ χ²(d₁) and X₂ ~ χ²(d₂) are independent, then the ratio:

This is the mathematical definition of the F-distribution. It follows directly that if you divide a chi-square random variable by its degrees of freedom, you get a scaled chi-square — and the ratio of two such scaled variables follows the F-distribution.

The Connection to the t-Distribution

If T follows a t-distribution with ν degrees of freedom, then T² follows an F-distribution with 1 and ν degrees of freedom:

This means a two-tailed t-test is a special case of the F-test. When comparing just two groups in ANOVA, the resulting F-statistic equals the square of the t-statistic you'd get from a two-sample t-test. The two approaches will always give the same p-value. To learn more about when to use each, see the hypothesis testing guide.

Real-World Applications of the F-Distribution

The F-distribution appears in any analysis that compares variation across groups or tests model fit. Here are the settings where you're most likely to encounter it.

F-Distribution Calculator

Enter your F-statistic and degrees of freedom to find the p-value. The calculator uses a numerical approximation of the regularized incomplete beta function — the same method used by R's pf() function.

For convenient online access to a full F-table with critical values, visit the F-distribution table. The ANOVA calculator at statisticsfundamentals.com/calculators/anova/ performs the complete sum-of-squares calculation from raw data.

F-Distribution Cheat Sheet

The Variance → F Decision Chain

Common Misconceptions

Frequently Asked Questions

Computing F-Tests in Statistical Software

Every major statistics package handles the F-distribution and ANOVA natively. Here's how to run a one-way ANOVA in the most common environments.

R

# One-way ANOVA in R
model <- aov(score ~ group, data = mydata)
summary(model)

# F-distribution p-value from F-statistic
pf(35.01, df1 = 2, df2 = 12, lower.tail = FALSE)

# Critical value at alpha = 0.05
qf(0.95, df1 = 2, df2 = 12)
          

Python (SciPy)

from scipy import stats

# One-way ANOVA
f_stat, p_value = stats.f_oneway(group1, group2, group3)

# F-distribution p-value
p = 1 - stats.f.cdf(35.01, dfn=2, dfd=12)

# Critical value at alpha = 0.05
crit = stats.f.ppf(0.95, dfn=2, dfd=12)
          

Excel

In Excel, use =F.DIST.RT(f_stat, d1, d2) to get the right-tail p-value, or =F.INV.RT(alpha, d1, d2) for the critical value.

Sources and Further Reading

• Fisher, R. A. (1924) — "On a distribution yielding the error functions of several well known statistics." Proceedings of the International Congress of Mathematics, Toronto, 2, 805–813. Original paper introducing what later became the F-distribution.
• Montgomery, D.C. (2017) — Design and Analysis of Experiments, 9th ed. Wiley. The standard graduate textbook for ANOVA and F-test methodology.
• NIST/SEMATECH e-Handbook of Statistical Methods — Section 7.4: Comparisons based on data from two processes. itl.nist.gov/div898/handbook/
• OpenStax Statistics — Chapter 13: F Distribution and One-Way ANOVA. Free, peer-reviewed textbook available at openstax.org
• Hogg, R.V., McKean, J.W., & Craig, A.T. (2018) — Introduction to Mathematical Statistics, 8th ed. Pearson. Provides the mathematical derivation of the F-distribution from first principles.
• SciPy Reference Guide — scipy.stats.f: F continuous random variable. docs.scipy.org