{
  "id": "conditional-expectation",
  "title": "Conditional Expectation",
  "lang": "en",
  "source_lang": "en",
  "summary": "Conditional expectation is the mean of a random variable given specified information, defined rigorously by measurability and integral-preservation conditions.",
  "category": "mathematics",
  "subcategory": "probability-statistics",
  "type": "concept",
  "aliases": [
    "Conditional expected value",
    "Conditional mean"
  ],
  "generated_at": "2026-10-04",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/conditional-expectation/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/conditional-expectation/",
      "json": "https://aiwiki.page/zh/conditional-expectation.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Conditional expectation is a fundamental concept in probability and statistics that describes the average value of a random variable given specified information. It generalizes the ordinary expected value: conditioning on an event produces a number, whereas conditioning on another random variable generally produces a random variable whose value depends on the observed information. It provides a framework for averaging within subpopulations, predicting unobserved quantities, and describing uncertainty as information accumulates. (ocw.mit.edu)"
    },
    {
      "heading": "Elementary definitions",
      "text": "For an integrable random variable \\(X\\) and an event \\(A\\) with \\(P(A)>0\\), the conditional expectation is\n\n\\[E[X\\mid A]=\\frac{E[X\\mathbf 1_A]}{P(A)},\\]\n\nwhere \\(\\mathbf 1_A\\) equals one on \\(A\\) and zero elsewhere. For discrete \\(X\\), this becomes\n\n\\[E[X\\mid A]=\\sum_x xP(X=x\\mid A).\\]\n\nThus the average uses probabilities adjusted to the specified event rather than the original probabilities. (ocw.mit.edu)\n\nIf \\(Y\\) is discrete, define\n\n\\[m(y)=E[X\\mid Y=y] =\\sum_x xP(X=x\\mid Y=y)\\]\n\nfor values with \\(P(Y=y)>0\\). The expression \\(E[X\\mid Y]=m(Y)\\) denotes a random variable, while \\(E[X\\mid Y=y]\\) denotes its numerical value at a specified observation. This distinction is essential: the conditional mean changes with \\(Y\\), even though the unconditional mean is constant. (ocw.mit.edu)\n\nFor real-valued \\(X,Y\\) having a joint distribution with density \\(f_{X,Y}\\), the corresponding formula uses a conditional density:\n\n\\[m(y)=\\int_{\\mathbb R}x\\,f_{X\\mid Y}(x\\mid y)\\,dx, \\qquad f_{X\\mid Y}(x\\mid y)=\\frac{f_{X,Y}(x,y)}{f_Y(y)},\\]\n\nwhere \\(f_Y(y)>0\\) and the integral exists. This is not ordinary division by \\(P(Y=y)\\), which is zero for continuously distributed \\(Y\\). (live.ocw.mit.edu)"
    },
    {
      "heading": "Measure-theoretic definition",
      "text": "The general definition belongs to measure theory. Let \\((\\Omega,\\mathcal F,P)\\) be a probability space, let \\(E|X|<\\infty\\), and let \\(\\mathcal G\\subseteq\\mathcal F\\) be a sigma-algebra representing the available information. A conditional expectation \\(Z=E[X\\mid\\mathcal G]\\) is an integrable, \\(\\mathcal G\\)-measurable random variable satisfying\n\n\\[\\int_A Z\\,dP=\\int_A X\\,dP \\quad\\text{for every }A\\in\\mathcal G.\\]\n\nThese integral identities preserve the averages detectable using that information. Measurability ensures that \\(Z\\) does not depend on information outside \\(\\mathcal G\\). (math.ucdavis.edu)\n\nExistence follows from the Radon–Nikodym theorem: for nonnegative integrable \\(X\\), the measure \\(A\\mapsto E[X\\mathbf 1_A]\\) on \\(\\mathcal G\\) has a Radon–Nikodym derivative relative to \\(P\\). General integrable \\(X\\) is handled through its positive and negative parts. Conditional expectations are unique almost surely, meaning that versions may differ on sets of probability zero. (math.ucdavis.edu)\n\nConditioning on \\(Y\\) means conditioning on its generated sigma-algebra \\(\\sigma(Y)\\). For real-valued \\(Y\\), a version can be written \\(m(Y)\\), with \\(m\\) measurable. Its values at individual observations of probability zero need not be uniquely determined; additional regularity may select a convenient version. (math.ucdavis.edu)"
    },
    {
      "heading": "Principal properties",
      "text": "For integrable variables, conditional expectation is linear and order-preserving. It leaves \\(\\mathcal G\\)-measurable variables unchanged. If \\(X\\) is independent of \\(\\mathcal G\\), then\n\n\\[E[X\\mid\\mathcal G]=E[X] \\quad\\text{almost surely}.\\]\n\nIndependence is sufficient for this identity, but the identity alone does not establish independence. (math.ucdavis.edu)\n\nThe “taking out what is known” rule states that, for bounded \\(\\mathcal G\\)-measurable \\(W\\),\n\n\\[E[WX\\mid\\mathcal G]=W E[X\\mid\\mathcal G].\\]\n\nThe tower property states that if \\(\\mathcal H\\subseteq\\mathcal G\\), then\n\n\\[E[E[X\\mid\\mathcal G]\\mid\\mathcal H] =E[X\\mid\\mathcal H].\\]\n\nAveraging a finer-information estimate using coarser information therefore produces the coarser-information estimate directly. Its special case, the law of total expectation, is\n\n\\[E[E[X\\mid\\mathcal G]]=E[X].\\]\n\nThese identities hold almost surely where appropriate. (people.math.wisc.edu)\n\nConditional Jensen’s inequality gives\n\n\\[\\varphi(E[X\\mid\\mathcal G]) \\leq E[\\varphi(X)\\mid\\mathcal G]\\]\n\nfor a convex function \\(\\varphi\\), assuming the relevant expectations exist. In particular, conditioning cannot increase the expected absolute magnitude:\n\n\\[E|E[X\\mid\\mathcal G]|\\leq E|X|.\\]\n\n(samuel-drapeau.info)"
    },
    {
      "heading": "Prediction and geometric interpretation",
      "text": "When \\(E[X^2]<\\infty\\), conditional expectation is the orthogonal projection of \\(X\\) onto the closed subspace of \\(\\mathcal G\\)-measurable variables in the Hilbert space \\(L^2\\). Its residual satisfies\n\n\\[E[(X-E[X\\mid\\mathcal G])W]=0\\]\n\nfor every square-integrable, \\(\\mathcal G\\)-measurable \\(W\\). Orthogonality here uses the inner product \\(E[UV]\\). (samuel-drapeau.info)\n\nConsequently, \\(E[X\\mid\\mathcal G]\\) minimizes mean squared error among square-integrable predictions based on \\(\\mathcal G\\). For observed predictors \\(Y\\), the minimizing prediction is \\(m(Y)=E[X\\mid Y]\\). Unlike linear regression, this optimization does not restrict predictions to linear functions; the optimal conditional mean may be nonlinear. (ocw.mit.edu)"
    },
    {
      "heading": "Variance decomposition and stochastic processes",
      "text": "For square-integrable \\(X\\), define conditional variance by\n\n\\[\\operatorname{Var}(X\\mid\\mathcal G) =E[(X-E[X\\mid\\mathcal G])^2\\mid\\mathcal G].\\]\n\nThe law of total variance then gives\n\n\\[\\operatorname{Var}(X) =E[\\operatorname{Var}(X\\mid\\mathcal G)] +\\operatorname{Var}(E[X\\mid\\mathcal G]).\\]\n\nIt separates average variability remaining within information-defined groups from variability between their conditional means. (ocw.mit.edu)\n\nIn a stochastic process, an increasing filtration \\((\\mathcal F_t)\\) represents information available over time. An integrable adapted process \\(M_t\\) is a martingale when\n\n\\[E[M_t\\mid\\mathcal F_s]=M_s,\\qquad s\\leq t.\\]\n\nFor an integrable terminal quantity \\(Z\\), the process \\(M_t=E[Z\\mid\\mathcal F_t]\\) satisfies this identity by the tower property, expressing successive estimates of the same quantity as information grows. (people.math.wisc.edu)"
    }
  ],
  "links": [
    "probability",
    "statistics",
    "random-variable",
    "expected-value",
    "joint-probability-distribution",
    "probability-density-function",
    "measure-theory",
    "probability-space",
    "sigma-algebra",
    "lebesgue-integral",
    "statistical-independence",
    "law-of-total-expectation",
    "jensens-inequality",
    "convex-function",
    "hilbert-space",
    "inner-product",
    "mean-squared-error",
    "linear-regression",
    "variance",
    "stochastic-process",
    "filtration-probability",
    "martingale"
  ],
  "sources": [
    {
      "title": "ocw.mit.edu/courses/6-041sc-probabilistic-systems-analysis-and-applied-probabil…",
      "url": "https://ocw.mit.edu/courses/6-041sc-probabilistic-systems-analysis-and-applied-probability-fall-2013/P7a4bjE6Crk_transcript.pdf"
    },
    {
      "title": "ocw.mit.edu/courses/6-042j-mathematics-for-computer-science-fall-2010/93cad640c…",
      "url": "https://ocw.mit.edu/courses/6-042j-mathematics-for-computer-science-fall-2010/93cad640cf3ed0b23ef70688f452d4d5_MIT6_042JF10_notes.pdf"
    },
    {
      "title": "live.ocw.mit.edu/courses/14-30-introduction-to-statistical-methods-in-economics…",
      "url": "https://live.ocw.mit.edu/courses/14-30-introduction-to-statistical-methods-in-economics-spring-2009/b5f0c35be0af22f4c8948e7db00ac9fa_MIT14_30s09_lec16.pdf"
    },
    {
      "title": "math.ucdavis.edu/~romik/data/uploads/notes/lecturenotes235b.pdf",
      "url": "https://www.math.ucdavis.edu/~romik/data/uploads/notes/lecturenotes235b.pdf"
    },
    {
      "title": "people.math.wisc.edu/~dfanderson/635S12/Lectures/635_Ch4_Lec.pdf",
      "url": "https://people.math.wisc.edu/~dfanderson/635S12/Lectures/635_Ch4_Lec.pdf"
    },
    {
      "title": "samuel-drapeau.info/SP_Lecture/lecture/03-Integration/033-radon-nikodym-cond-ex…",
      "url": "https://www.samuel-drapeau.info/SP_Lecture/lecture/03-Integration/033-radon-nikodym-cond-exp/"
    },
    {
      "title": "ocw.mit.edu/courses/res-6-012-introduction-to-probability-spring-2018/55e1f9855…",
      "url": "https://ocw.mit.edu/courses/res-6-012-introduction-to-probability-spring-2018/55e1f98550110cea4b1980ec9df6224d_HL7qwWvON4.pdf"
    }
  ],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}