{
  "id": "f-distribution",
  "title": "F Distribution",
  "lang": "en",
  "source_lang": "en",
  "summary": "The F distribution is a continuous probability distribution of a ratio of independent, scaled chi-squared variables, widely used in variance comparisons, analysis of variance, and regression tests.",
  "category": "mathematics",
  "subcategory": "probability-statistics",
  "type": "concept",
  "aliases": [
    "Fisher–Snedecor distribution",
    "Snedecor's F distribution"
  ],
  "generated_at": "2026-11-04",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/f-distribution/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/f-distribution/",
      "json": "https://aiwiki.page/zh/f-distribution.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "The F distribution is a continuous probability distribution obtained by dividing two independent chi-squared random variables, each normalized by its degrees of freedom. It is a fundamental sampling distribution in statistics, especially for comparing variances, conducting analysis of variance, and testing restrictions in linear regression. The ordinary F distribution is also called the central F distribution, distinguishing it from the noncentral F distribution that arises under many alternative hypotheses."
    },
    {
      "heading": "Definition",
      "text": "Let \\(U\\) and \\(V\\) be random variables such that\n\n\\[U\\sim\\chi^2_{d_1}, \\qquad V\\sim\\chi^2_{d_2},\\]\n\nwith statistical independence between them. Then\n\n\\[X=\\frac{U/d_1}{V/d_2}\\]\n\nhas an F distribution with numerator degrees of freedom \\(d_1\\) and denominator degrees of freedom \\(d_2\\), written\n\n\\[X\\sim F_{d_1,d_2}.\\]\n\nBoth parameters may be any positive real numbers, although they are usually positive integers in elementary applications.\n\nAn F-distributed variable is positive with probability one. The order of its parameters matters: \\(F_{d_1,d_2}\\) and \\(F_{d_2,d_1}\\) are generally different distributions. Independence is essential; a ratio of dependent scaled chi-squared variables need not have an F distribution."
    },
    {
      "heading": "Density and distribution function",
      "text": "For \\(x>0\\), the probability density function is\n\n\\[f(x;d_1,d_2) = \\frac{\\Gamma\\!\\left((d_1+d_2)/2\\right)} {\\Gamma(d_1/2)\\Gamma(d_2/2)} \\left(\\frac{d_1}{d_2}\\right)^{d_1/2} x^{d_1/2-1} \\left(1+\\frac{d_1}{d_2}x\\right)^{-(d_1+d_2)/2},\\]\n\nwhere \\(\\Gamma\\) denotes the gamma function. The density is zero for \\(x<0\\).\n\nThe cumulative distribution function can be expressed through the regularized incomplete beta function:\n\n\\[P(X\\le x) = I_{\\frac{d_1x}{d_2+d_1x}} \\left(\\frac{d_1}{2},\\frac{d_2}{2}\\right), \\qquad x>0,\\]\n\nwhere\n\n\\[I_z(a,b) = \\frac{\\displaystyle\\int_0^z t^{a-1}(1-t)^{b-1}\\,dt} {\\displaystyle\\int_0^1 t^{a-1}(1-t)^{b-1}\\,dt}.\\]\n\nThis representation connects the F distribution to beta-distribution calculations and is important for numerical evaluation of probabilities and critical values."
    },
    {
      "heading": "Shape and moments",
      "text": "The F distribution is generally right-skewed. Its shape depends on both degrees-of-freedom parameters:\n\n- If \\(d_1<2\\), the density diverges as \\(x\\) approaches zero from above.\n- If \\(d_1=2\\), the density approaches a finite, positive value at zero and decreases thereafter.\n- If \\(d_1>2\\), the density rises to an interior mode and then decreases.\n\nFor \\(d_1>2\\), the mode is\n\n\\[\\operatorname{mode}(X) = \\frac{d_2(d_1-2)}{d_1(d_2+2)}.\\]\n\nThe mean exists only when \\(d_2>2\\), in which case\n\n\\[\\operatorname{E}[X] = \\frac{d_2}{d_2-2}.\\]\n\nIn particular, the mean is not generally one, even though the distribution is constructed from a ratio of two variables whose normalized means are each one.\n\nThe variance exists only when \\(d_2>4\\), and is\n\n\\[\\operatorname{Var}(X) = \\frac{2d_2^2(d_1+d_2-2)} {d_1(d_2-2)^2(d_2-4)}.\\]\n\nMore generally, for real \\(r\\),\n\n\\[\\operatorname{E}[X^r] = \\left(\\frac{d_2}{d_1}\\right)^r \\frac{ \\Gamma(d_1/2+r)\\Gamma(d_2/2-r) }{ \\Gamma(d_1/2)\\Gamma(d_2/2) },\\]\n\nprovided\n\n\\[-\\frac{d_1}{2}<r<\\frac{d_2}{2}.\\]\n\nThe denominator degrees of freedom determine the existence of positive moments. For large \\(x\\), the density is proportional asymptotically to \\(x^{-d_2/2-1}\\); small denominator degrees of freedom therefore produce particularly heavy right tails."
    },
    {
      "heading": "Relationships with other distributions",
      "text": "Reciprocal identity\n\nReversing the numerator and denominator gives\n\n\\[X\\sim F_{d_1,d_2} \\quad\\Longrightarrow\\quad \\frac{1}{X}\\sim F_{d_2,d_1}.\\]\n\nIf \\(q_p(d_1,d_2)\\) denotes the \\(p\\)-quantile of \\(F_{d_1,d_2}\\), then\n\n\\[q_p(d_1,d_2) = \\frac{1}{q_{1-p}(d_2,d_1)}, \\qquad 0<p<1.\\]\n\nThis identity is useful for converting lower-tail critical values into upper-tail critical values.\n\nBeta distribution\n\nIf \\(X\\sim F_{d_1,d_2}\\), then\n\n\\[Y=\\frac{d_1X}{d_2+d_1X} \\sim \\operatorname{Beta}\\left(\\frac{d_1}{2},\\frac{d_2}{2}\\right).\\]\n\nConversely, if \\(Y\\) has this beta distribution, then\n\n\\[X=\\frac{d_2Y}{d_1(1-Y)}\\]\n\nhas the corresponding F distribution.\n\nStudent's t distribution\n\nIf \\(T\\) has Student's t distribution with \\(\\nu\\) degrees of freedom, then\n\n\\[T^2\\sim F_{1,\\nu}.\\]\n\nConsequently, in the classical normal linear model, a two-sided t-test of a single coefficient is equivalent to the corresponding F-test with one numerator degree of freedom.\n\nLimiting distributions\n\nFor fixed \\(d_1\\), as \\(d_2\\) tends to infinity,\n\n\\[F_{d_1,d_2} \\xrightarrow{d} \\frac{\\chi^2_{d_1}}{d_1}.\\]\n\nFor fixed \\(d_2\\), as \\(d_1\\) tends to infinity,\n\n\\[F_{d_1,d_2} \\xrightarrow{d} \\frac{d_2}{\\chi^2_{d_2}}.\\]\n\nWhen both degrees of freedom tend to infinity, the distribution concentrates around one."
    },
    {
      "heading": "Statistical applications",
      "text": "An F-test is a hypothesis test whose test statistic follows an F distribution under the null hypothesis, either exactly under specified assumptions or approximately. Different F-tests use different statistics and assumptions; the name does not identify a single testing procedure.\n\nComparing two population variances\n\nConsider two independent samples of sizes \\(n_1\\) and \\(n_2\\) from normal populations with variances \\(\\sigma_1^2\\) and \\(\\sigma_2^2\\). Let \\(S_1^2\\) and \\(S_2^2\\) be the sample variances calculated using denominators \\(n_1-1\\) and \\(n_2-1\\). Then\n\n\\[\\frac{S_1^2/\\sigma_1^2}{S_2^2/\\sigma_2^2} \\sim F_{n_1-1,n_2-1}.\\]\n\nUnder the hypothesis \\(\\sigma_1^2=\\sigma_2^2\\),\n\n\\[\\frac{S_1^2}{S_2^2} \\sim F_{n_1-1,n_2-1}.\\]\n\nA two-sided test uses both tails because either unusually large or unusually small ratios can contradict equality.\n\nThe same relationship yields an exact confidence interval for the variance ratio. Write \\(R=S_1^2/S_2^2\\), and let \\(q_p\\) denote the \\(p\\)-quantile of \\(F_{n_1-1,n_2-1}\\). An equal-tailed interval with confidence level \\(1-\\alpha\\) is\n\n\\[\\left[ \\frac{R}{q_{1-\\alpha/2}}, \\frac{R}{q_{\\alpha/2}} \\right].\\]\n\nAnalysis of variance\n\nIn a one-way analysis of variance with \\(k\\) independent groups and \\(N\\) total observations, the classical statistic is\n\n\\[F= \\frac{SS_{\\text{between}}/(k-1)} {SS_{\\text{within}}/(N-k)}.\\]\n\nHere \\(SS_{\\text{between}}\\) measures variation among group sample means, weighted by group size, and \\(SS_{\\text{within}}\\) measures variation around the respective group means.\n\nUnder equal population means, independent normally distributed errors, and a common error variance,\n\n\\[F\\sim F_{k-1,N-k}.\\]\n\nLarge values provide evidence against equality of all group means. Rejection establishes that the means are not all equal; it does not identify which groups differ. Follow-up comparisons constitute additional inference and may involve multiple-testing considerations.\n\nNested regression models\n\nFor two nested linear regression models fitted to the same \\(n\\) observations by ordinary least squares, let:\n\n- \\(RSS_R\\) be the restricted model's residual sum of squares;\n- \\(RSS_U\\) be the unrestricted model's residual sum of squares;\n- \\(q\\) be the number of independent restrictions;\n- \\(p\\) be the number of freely estimated regression coefficients in the unrestricted model.\n\nThe statistic\n\n\\[F= \\frac{(RSS_R-RSS_U)/q} {RSS_U/(n-p)}\\]\n\nhas an \\(F_{q,n-p}\\) distribution under the restrictions, provided the classical normal, independent, constant-variance error assumptions hold and the model matrices have the required ranks.\n\nThis formulation includes tests of whether several regression coefficients are jointly zero and the overall significance test for a regression with an intercept.\n\nTail probabilities\n\nFor an upper-tail F-test with observed value \\(f_{\\mathrm{obs}}\\), the p-value is\n\n\\[p=P(F_{d_1,d_2}\\ge f_{\\mathrm{obs}}).\\]\n\nThe rejection threshold at significance level \\(\\alpha\\) is \\(q_{1-\\alpha}(d_1,d_2)\\). Two-sided variance tests require a specified two-tail procedure rather than automatic use of this upper-tail rule."
    },
    {
      "heading": "Noncentral F distribution",
      "text": "The noncentral F distribution extends the construction by allowing the numerator to have a noncentral chi-squared distribution:\n\n\\[X= \\frac{U/d_1}{V/d_2}, \\qquad U\\sim\\chi^2_{d_1}(\\lambda), \\qquad V\\sim\\chi^2_{d_2},\\]\n\nwhere \\(U\\) and \\(V\\) are independent and \\(\\lambda\\ge0\\) is the noncentrality parameter. It is commonly written \\(F_{d_1,d_2}(\\lambda)\\). Setting \\(\\lambda=0\\) gives the central F distribution.\n\nThe noncentral distribution describes many F-statistics when an alternative hypothesis is true. It is used to calculate statistical power and determine sample sizes for specified alternatives.\n\nFor example, in a fixed-effects one-way normal model with common variance \\(\\sigma^2\\),\n\n\\[\\lambda = \\frac{\\sum_{i=1}^{k}n_i(\\mu_i-\\bar{\\mu}_w)^2}{\\sigma^2}, \\qquad \\bar{\\mu}_w = \\frac{\\sum_i n_i\\mu_i}{\\sum_i n_i}.\\]\n\nHere \\(n_i\\) and \\(\\mu_i\\) are the size and population mean of group \\(i\\). The parameter combines the separation of group means, their sample sizes, and the error variance."
    },
    {
      "heading": "Assumptions and limitations",
      "text": "The exact F laws used in classical procedures arise from independent chi-squared quantities with appropriate scale factors. In many applications, this structure follows from a normal model and an orthogonal decomposition of sums of squares.\n\nImportant limitations include:\n\n- Sensitivity of variance-ratio tests to nonnormality. The classical two-sample variance test can have substantially incorrect rejection rates for nonnormal populations.\n- Unequal error variances. Classical analysis-of-variance and regression F-tests need not retain their stated null distributions when errors have unequal variances.\n- Dependent observations. Repeated measurements, clustered data, or serially correlated errors generally invalidate the simplest independent-error formulas.\n- Exact versus approximate reference distributions. Some procedures use F approximations with adjusted, noninteger degrees of freedom. Their calibration is distinct from the exact chi-squared-ratio derivation.\n\nA statistically significant F-test concerns the particular joint hypothesis being tested. It does not, by itself, measure practical importance or establish causation."
    },
    {
      "heading": "Historical background",
      "text": "The distribution is closely associated with Ronald A. Fisher and the development of variance-based statistical inference in the early twentieth century. George W. Snedecor introduced the designation F in Fisher's honor and helped establish its practical use through statistical tables and agricultural applications. The alternative name Fisher–Snedecor distribution reflects these contributions."
    }
  ],
  "links": [
    "probability-distribution",
    "chi-squared-distribution",
    "degrees-of-freedom",
    "sampling-distribution",
    "statistics",
    "analysis-of-variance",
    "linear-regression",
    "statistical-independence",
    "probability-density-function",
    "gamma-function",
    "cumulative-distribution-function",
    "expected-value",
    "variance",
    "quantile",
    "beta-distribution",
    "students-t-distribution",
    "statistical-hypothesis-testing",
    "test-statistic",
    "null-hypothesis",
    "confidence-interval",
    "multiple-testing",
    "ordinary-least-squares",
    "p-value",
    "significance-level",
    "alternative-hypothesis",
    "statistical-power",
    "sample-size",
    "normal-distribution",
    "ronald-fisher"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}