{
  "id": "convergence-in-distribution",
  "title": "Convergence in Distribution",
  "lang": "en",
  "source_lang": "en",
  "summary": "Convergence in distribution is convergence of probability laws, characterized by convergence of distribution functions at continuity points or of expectations of bounded continuous functions.",
  "category": "mathematics",
  "subcategory": "probability-statistics",
  "type": "concept",
  "aliases": [
    "Convergence in law",
    "Weak convergence of probability distributions",
    "Distributional convergence"
  ],
  "generated_at": "2026-11-05",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/convergence-in-distribution/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/convergence-in-distribution/",
      "json": "https://aiwiki.page/zh/convergence-in-distribution.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Convergence in distribution is a mode of convergence in probability theory in which the probability distributions of a sequence of random variables approach a limiting distribution. It describes the limiting probabilities of outcomes rather than requiring the random variables themselves to become close. For real-valued random variables, it means that their cumulative distribution functions converge at every continuity point of the limiting distribution function. This notion is fundamental to limit theorems and to distributional approximations in statistics."
    },
    {
      "heading": "Definition",
      "text": "Let \\(X_1,X_2,\\ldots\\) and \\(X\\) be real-valued random variables, with cumulative distribution functions\n\n\\[F_n(x)=\\mathbb P(X_n\\leq x), \\qquad F(x)=\\mathbb P(X\\leq x).\\]\n\nThe sequence \\(X_n\\) converges in distribution to \\(X\\) if\n\n\\[F_n(x)\\longrightarrow F(x)\\]\n\nfor every \\(x\\in\\mathbb R\\) at which \\(F\\) is continuous. Standard notation includes\n\n\\[X_n\\xrightarrow{d}X, \\qquad X_n\\Rightarrow X, \\qquad X_n\\xrightarrow{\\mathcal D}X.\\]\n\nOnly the distributions matter. The variables need not be defined on the same probability space, and replacing any of them with another variable having the same distribution leaves the convergence statement unchanged. Equivalently, if \\(\\mu_n\\) and \\(\\mu\\) denote their laws, one writes\n\n\\[\\mu_n\\Rightarrow\\mu.\\]\n\nThis is called weak convergence of probability measures. Here, “weak” refers to testing measures against suitable functions; it does not mean convergence to a generalized function in distribution theory.\n\nWhy continuity points are required\n\nSuppose \\(X_n=1/n\\) deterministically and \\(X=0\\). Then \\(X_n\\Rightarrow X\\), but\n\n\\[F_n(0)=0 \\quad\\text{for all }n, \\qquad F(0)=1.\\]\n\nThe limiting distribution has an atom at zero, so its distribution function is discontinuous there. Requiring convergence at every real number would incorrectly exclude this sequence.\n\nIf \\(F\\) is continuous everywhere, however, convergence in distribution implies the stronger statement\n\n\\[\\sup_{x\\in\\mathbb R}|F_n(x)-F(x)|\\longrightarrow0.\\]\n\nThus, for a continuous limiting distribution function, the convergence of distribution functions is uniform."
    },
    {
      "heading": "Equivalent characterizations",
      "text": "Bounded continuous test functions\n\nFor real-valued random variables, convergence in distribution is equivalent to\n\n\\[\\mathbb E[f(X_n)]\\longrightarrow\\mathbb E[f(X)]\\]\n\nfor every bounded continuous function\n\\(f:\\mathbb R\\to\\mathbb R\\).\n\nThis formulation expresses convergence through averaged observations. Boundedness ensures that the relevant expectations exist; continuity prevents a test function from detecting an unstable distinction at a boundary carrying limiting probability.\n\nIt also extends the definition to random elements in more general spaces. If \\(S\\) is a metric space and \\(\\mu_n,\\mu\\) are probability measures on its Borel sigma-algebra, weak convergence is defined by\n\n\\[\\int_S f\\,d\\mu_n\\longrightarrow\\int_S f\\,d\\mu\\]\n\nfor every bounded continuous real-valued function on \\(S\\).\n\nFor vectors in Euclidean space, convergence of multivariate distribution functions at every continuity point of the limiting distribution function gives an equivalent criterion.\n\nPortmanteau theorem\n\nThe Portmanteau theorem collects equivalent formulations of weak convergence. For Borel probability measures on a metric space, \\(\\mu_n\\Rightarrow\\mu\\) is equivalent to each of the following conditions:\n\n- For every closed set \\(C\\),\n\n\\[\\limsup_{n\\to\\infty}\\mu_n(C)\\leq\\mu(C).\\]\n- For every open set \\(G\\),\n\n\\[\\liminf_{n\\to\\infty}\\mu_n(G)\\geq\\mu(G).\\]\n- For every Borel set \\(A\\) whose boundary satisfies \\(\\mu(\\partial A)=0\\),\n\n\\[\\mu_n(A)\\longrightarrow\\mu(A).\\]\n\nA set satisfying the boundary condition is called a continuity set of \\(\\mu\\). This qualification is essential: weak convergence does not imply convergence of probabilities for every measurable set.\n\nFor example, for real numbers \\(a<b\\),\n\n\\[\\mathbb P(a<X_n\\leq b)\\longrightarrow \\mathbb P(a<X\\leq b)\\]\n\nwhenever \\(X\\) has no atoms at \\(a\\) or \\(b\\).\n\nCharacteristic functions\n\nThe characteristic function of a real-valued random variable is\n\n\\[\\varphi_X(t)=\\mathbb E[e^{itX}], \\qquad t\\in\\mathbb R.\\]\n\nLévy’s continuity theorem states that\n\n\\[X_n\\Rightarrow X \\quad\\Longleftrightarrow\\quad \\varphi_{X_n}(t)\\longrightarrow\\varphi_X(t) \\quad\\text{for every }t\\in\\mathbb R.\\]\n\nMore generally, if characteristic functions converge pointwise to a function \\(\\varphi\\) that is continuous at zero, then \\(\\varphi\\) is the characteristic function of a probability distribution, and the corresponding laws converge weakly to that distribution. This criterion is especially useful for sums of independent variables because their characteristic functions multiply."
    },
    {
      "heading": "Relationship to other modes of convergence",
      "text": "For variables defined on a common probability space, the principal implications are\n\n\\[X_n\\longrightarrow X\\text{ almost surely} \\quad\\Longrightarrow\\quad X_n\\xrightarrow{p}X \\quad\\Longrightarrow\\quad X_n\\xrightarrow{d}X.\\]\n\nAlmost sure convergence concerns convergence outside a probability-zero set of outcomes. Convergence in probability requires\n\n\\[\\mathbb P(|X_n-X|>\\varepsilon)\\longrightarrow0\\]\n\nfor every \\(\\varepsilon>0\\). Convergence in distribution is weaker: it constrains only the individual laws, not how the variables are coupled.\n\nFor example, let \\(X\\) have the Bernoulli distribution with parameter \\(1/2\\), and set \\(X_n=1-X\\) for every \\(n\\). Each \\(X_n\\) has exactly the same distribution as \\(X\\), so \\(X_n\\Rightarrow X\\). Nevertheless,\n\n\\[\\mathbb P(|X_n-X|=1)=1,\\]\n\nand convergence in probability to \\(X\\) fails.\n\nAn important exception occurs when the limiting distribution is concentrated at a constant \\(c\\):\n\n\\[X_n\\xrightarrow{d}c \\quad\\Longleftrightarrow\\quad X_n\\xrightarrow{p}c.\\]\n\nA constant can be regarded as a random variable on any probability space supporting \\(X_n\\)."
    },
    {
      "heading": "Fundamental examples",
      "text": "Central limit theorem\n\nA principal example is the central limit theorem. If \\(Y_1,Y_2,\\ldots\\) are independent and identically distributed with mean \\(m\\) and finite, positive variance \\(\\sigma^2\\), then\n\n\\[\\frac{\\sum_{k=1}^nY_k-nm}{\\sigma\\sqrt n} \\xrightarrow{d} Z, \\qquad Z\\sim N(0,1).\\]\n\nThe limiting normal distribution provides an approximation to probabilities involving the standardized sum. If the original variables are discrete, the standardized sums remain discrete for every finite \\(n\\); convergence in distribution does not require the approximating and limiting laws to have the same type.\n\nPoisson limit\n\nSuppose \\(X_n\\) has the binomial distribution with parameters \\(n\\) and \\(p_n\\), where\n\n\\[np_n\\longrightarrow\\lambda, \\qquad 0<\\lambda<\\infty.\\]\n\nThen\n\n\\[X_n\\xrightarrow{d}X, \\qquad X\\sim\\operatorname{Poisson}(\\lambda).\\]\n\nThis gives the Poisson distribution as a limit for counts of many independent events whose individual probabilities become small.\n\nDiscrete laws approaching a continuous law\n\nLet \\(X_n\\) assign probability \\(1/n\\) to each point\n\n\\[\\frac1n,\\frac2n,\\ldots,1.\\]\n\nThen \\(X_n\\Rightarrow U\\), where \\(U\\) has the uniform distribution on \\([0,1]\\). For every bounded continuous \\(f\\),\n\n\\[\\mathbb E[f(X_n)] =\\frac1n\\sum_{k=1}^n f(k/n) \\longrightarrow \\int_0^1f(x)\\,dx.\\]\n\nYet if \\(A=\\mathbb Q\\cap[0,1]\\), then\n\n\\[\\mathbb P(X_n\\in A)=1, \\qquad \\mathbb P(U\\in A)=0.\\]\n\nThis illustrates why weak convergence cannot guarantee convergence of probabilities for arbitrary measurable sets."
    },
    {
      "heading": "Operations that preserve convergence",
      "text": "Continuous mapping theorem\n\nThe continuous mapping theorem states that if \\(X_n\\Rightarrow X\\) and \\(g\\) is continuous, then\n\n\\[g(X_n)\\Rightarrow g(X).\\]\n\nMore generally, a measurable \\(g\\) is sufficient if its set of discontinuities has probability zero under the law of \\(X\\).\n\nFor example,\n\n\\[X_n\\Rightarrow Z \\quad\\Longrightarrow\\quad X_n^2\\Rightarrow Z^2.\\]\n\nFor the reciprocal map, the corresponding conclusion requires that the limiting variable have no mass at zero, with an appropriate definition of the map there.\n\nSlutsky’s theorem\n\nSlutsky’s theorem states that if\n\n\\[X_n\\Rightarrow X, \\qquad Y_n\\xrightarrow{p}c,\\]\n\nwhere \\(c\\) is constant, then\n\n\\[X_n+Y_n\\Rightarrow X+c, \\qquad X_nY_n\\Rightarrow cX.\\]\n\nIf \\(c\\ne0\\), it also gives\n\n\\[\\frac{X_n}{Y_n}\\Rightarrow\\frac{X}{c},\\]\n\nprovided the ratio is defined, or assigned a value on events whose probabilities tend to zero.\n\nThis permits unknown scale parameters in asymptotic statistics to be replaced by consistent estimates.\n\nJoint convergence is a separate requirement\n\nMarginal convergence does not generally imply joint convergence. Suppose \\(X\\) takes values \\(1\\) and \\(-1\\) with equal probabilities, and define\n\n\\[X_n=X,\\qquad Y_n=(-1)^nX.\\]\n\nBoth marginal distributions are constant in \\(n\\), but the pair \\((X_n,Y_n)\\) alternates between two different joint distributions. In particular, \\(X_n+Y_n\\) alternates between having the law of \\(2X\\) and being identically zero.\n\nConsequently, applying a continuous function to several converging variables generally requires convergence of the whole vector, not merely convergence of each coordinate’s distribution."
    },
    {
      "heading": "Limitations",
      "text": "Moments need not converge\n\nConvergence in distribution does not imply convergence of means, variances, or other unbounded moments. Consider\n\n\\[X_n= \\begin{cases} n,&\\text{with probability }1/n,\\\\ 0,&\\text{with probability }1-1/n. \\end{cases}\\]\n\nThen \\(X_n\\to0\\) in probability and hence in distribution, but\n\n\\[\\mathbb E[X_n]=1\\]\n\nfor every \\(n\\).\n\nThe rare, large values disappear from the limiting distribution while retaining their contribution to the expectation. Additional control, such as uniform integrability, is needed to infer convergence of expectations.\n\nDensities need not converge or exist\n\nWeak convergence does not imply pointwise convergence of probability densities, and the limiting law need not possess a density. For example,\n\n\\[X_n\\sim N(0,1/n)\\]\n\nconverges in distribution to a point mass at zero. Conversely, the discrete uniform-grid example has a continuous limiting distribution despite having no densities with respect to Lebesgue measure at any finite stage.\n\nNo convergence rate is specified\n\nThe statement \\(X_n\\Rightarrow X\\) is asymptotic. It does not provide an error bound for a particular finite \\(n\\). Quantitative results, such as bounds on the difference between distribution functions, require additional assumptions and separate theorems."
    },
    {
      "heading": "Tightness and process limits",
      "text": "A family of probability measures on a metric space is tight if, for every \\(\\varepsilon>0\\), there is a compact set \\(K\\) such that every measure in the family assigns at least \\(1-\\varepsilon\\) probability to \\(K\\).\n\nFor a complete separable metric space, Prokhorov’s theorem identifies tightness with relative compactness under weak convergence: every sequence from a tight family has a weakly convergent subsequence. This gives a standard proof strategy—establish tightness, then show that every possible subsequential limit has the desired law.\n\nFor stochastic processes, convergence in distribution can be studied on a space of entire sample paths. The topology on that space is part of the convergence statement. Convergence of all finite-dimensional distributions alone generally does not imply weak convergence of the path-valued processes; suitable tightness conditions are also needed."
    },
    {
      "heading": "Role in statistical inference",
      "text": "Distributional limits describe the large-sample behavior of statistics. A common result for an estimator \\(\\widehat\\theta_n\\) is\n\n\\[\\sqrt n(\\widehat\\theta_n-\\theta) \\xrightarrow{d}N(0,V).\\]\n\nCombined with a consistent estimate of \\(V\\), this can justify approximate confidence intervals and hypothesis tests.\n\nThe limit concerns the centered and scaled estimation error. It does not mean that the unscaled estimator has a nondegenerate normal limit. Indeed, the displayed convergence implies that \\(\\widehat\\theta_n\\) converges in probability to \\(\\theta\\). The usefulness of any finite-sample approximation remains dependent on how closely the statistic’s actual distribution resembles its limiting law."
    }
  ],
  "links": [
    "probability",
    "probability-distribution",
    "statistics",
    "random-variable",
    "cumulative-distribution-function",
    "probability-space",
    "continuous-function",
    "expected-value",
    "metric-space",
    "borel-sigma-algebra",
    "euclidean-space",
    "closed-set",
    "open-set",
    "almost-surely",
    "convergence-in-probability",
    "bernoulli-distribution",
    "central-limit-theorem",
    "variance",
    "normal-distribution",
    "binomial-distribution",
    "poisson-distribution",
    "uniform-distribution",
    "joint-probability-distribution",
    "stochastic-process",
    "estimator"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}