{
  "id": "autocorrelation-function",
  "title": "Autocorrelation Function",
  "lang": "en",
  "source_lang": "en",
  "summary": "The autocorrelation function measures dependence between a process or signal and a time-shifted version of itself, linking temporal structure to frequency content.",
  "category": "mathematics",
  "subcategory": "probability-statistics",
  "type": "concept",
  "aliases": [
    "ACF",
    "serial correlation function"
  ],
  "generated_at": "2026-11-04",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/autocorrelation-function/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/autocorrelation-function/",
      "json": "https://aiwiki.page/zh/autocorrelation-function.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "The autocorrelation function describes how a quantity is related to itself at different separations in time, space, or another index. In time-series analysis, it measures the linear association between observations separated by a specified lag. In signal analysis, closely related definitions measure the overlap between a signal and a shifted copy of itself. These formulations make autocorrelation a central tool for identifying persistence, periodicity, characteristic scales, and departures from randomness.\n\nFor a stationary random process with finite, nonzero variance, the normalized autocorrelation at lag \\(k\\) is the covariance between observations \\(k\\) steps apart divided by their common variance. Other conventions use an uncentered expected product or an unnormalized integral, so the meaning of “autocorrelation” depends partly on the field. The distinction matters whenever a signal has a nonzero mean or its amplitude carries physical information.\n\nAutocorrelation connects the time-domain description of a process with its frequency-domain power spectrum. It also enters forecasting, model diagnostics, uncertainty estimation, signal detection, and the analysis of correlated simulations. Its interpretation requires care: zero autocorrelation does not generally imply independence, and an empirical autocorrelation pattern can arise from trends, changing regimes, or measurement procedures rather than stationary dependence."
    },
    {
      "heading": "Definitions and conventions",
      "text": "Let \\(\\{X_t\\}\\) be a real-valued stochastic process indexed by integer time. Suppose it has constant mean \\(\\mu\\), finite second moments, and covariance depending only on time separation. Its autocovariance function is\n\n\\[\\gamma(k)=\\operatorname{E}\\big[(X_{t+k}-\\mu)(X_t-\\mu)\\big],\\]\n\nwhere \\(\\operatorname{E}\\) denotes expectation. If \\(\\gamma(0)>0\\), its normalized autocorrelation function is\n\n\\[\\rho(k)=\\frac{\\gamma(k)}{\\gamma(0)}.\\]\n\nHere \\(\\gamma(0)\\) is the variance of each observation, and \\(k\\) is the lag. For example, \\(\\rho(1)\\) describes association between consecutive measurements, whereas \\(\\rho(12)\\) describes association between measurements twelve sampling intervals apart. The lag has no fixed physical duration until the sampling interval is specified.\n\nThe assumptions above define weak, or covariance, stationarity. They are weaker than requiring every finite-dimensional joint distribution to remain unchanged under time translation. For a nonstationary process, a two-time covariance can still be defined:\n\n\\[C(s,t)= \\operatorname{E}\\big[(X_s-\\mu_s)(X_t-\\mu_t)\\big],\\]\n\nbut it cannot generally be reduced to a function of \\(s-t\\). Its normalized counterpart uses the separate standard deviations at the two times. Calling this object a lag-only autocorrelation without additional assumptions discards potentially important information.\n\nFor complex-valued processes, conjugation is needed. A common convention is\n\n\\[\\gamma(k)= \\operatorname{E}\\big[(X_{t+k}-\\mu)\\overline{(X_t-\\mu)}\\big].\\]\n\nThis agrees with the real-valued definition when the observations are real. Some authors reverse the lag orientation or conjugate the other factor, which changes certain signs and complex conjugations but not the underlying information. For general complex processes, the covariance alone need not capture all second-order structure; a complementary covariance without conjugation may also be required.\n\nIn engineering, “autocorrelation” often means the uncentered function\n\n\\[R(k)=\\operatorname{E}[X_{t+k}\\overline{X_t}],\\]\n\nrather than its centered, normalized version. Under stationarity,\n\n\\[R(k)=\\gamma(k)+|\\mu|^2.\\]\n\nThus a nonzero mean contributes a constant component at every lag. Autocovariance has squared measurement units, while normalized autocorrelation is dimensionless.\n\nFor a deterministic finite-energy continuous-time signal \\(x(t)\\), another standard definition is\n\n\\[R_x(\\tau)= \\int_{-\\infty}^{\\infty} x(t+\\tau)\\overline{x(t)}\\,dt.\\]\n\nAn infinite-duration power signal instead calls for a suitable long-time average, when that limit exists. These deterministic definitions resemble ensemble expectations, but identifying a time average with a probabilistic expectation requires additional assumptions, usually involving ergodicity."
    },
    {
      "heading": "Structural properties and admissible functions",
      "text": "An autocorrelation function cannot be an arbitrary curve bounded between minus one and one. It must satisfy consistency conditions imposed by its origin as a covariance. For a real stationary process,\n\n\\[\\gamma(-k)=\\gamma(k), \\qquad \\rho(0)=1, \\qquad |\\rho(k)|\\leq 1.\\]\n\nFor complex processes, symmetry becomes Hermitian symmetry:\n\n\\[\\gamma(-k)=\\overline{\\gamma(k)}.\\]\n\nThe bound follows from the Cauchy–Schwarz inequality applied to the two centered random variables. Equality at a nonzero lag means that the corresponding variables are linearly dependent in the mean-square sense; it does not merely indicate a strong empirical association.\n\nThe central admissibility requirement is positive semidefiniteness. For any finite collection of times \\(t_1,\\ldots,t_m\\) and complex coefficients \\(a_1,\\ldots,a_m\\),\n\n\\[\\sum_{i=1}^{m}\\sum_{j=1}^{m} \\overline{a_i}a_j\\gamma(t_i-t_j)\\geq 0.\\]\n\nThis expression is the expectation of the squared absolute value of a suitable linear combination of centered observations. Consequently, every finite covariance matrix formed from the function must be positive semidefinite.\n\nFor consecutive integer times, these matrices have constant diagonals:\n\n\\[\\Gamma_m= \\begin{pmatrix} \\gamma(0)&\\gamma(-1)&\\cdots&\\gamma(1-m)\\\\ \\gamma(1)&\\gamma(0)&\\cdots&\\gamma(2-m)\\\\ \\vdots&\\vdots&\\ddots&\\vdots\\\\ \\gamma(m-1)&\\gamma(m-2)&\\cdots&\\gamma(0) \\end{pmatrix}.\\]\n\nSuch a Toeplitz matrix expresses translation invariance algebraically. Positive semidefiniteness is stronger than checking each individual lag. For example, the values \\(\\rho(1)=0.9\\) and \\(\\rho(2)=-0.9\\) both satisfy the elementary bound, but the resulting \\(3\\times3\\) correlation matrix has negative determinant and cannot represent a stationary process.\n\nConversely, a Hermitian positive-semidefinite function with finite value at zero can serve as a stationary covariance function. One construction uses a Gaussian process, whose finite-dimensional normal distributions are specified by a mean and compatible covariance matrices. This establishes that the positivity requirement is not merely necessary: it characterizes admissibility at the second-order level.\n\nAutocorrelation is insensitive to some transformations. If \\(Y_t=aX_t+b\\) with real \\(a\\neq0\\), then \\(Y_t\\) has the same normalized autocorrelation as \\(X_t\\), although its autocovariance is multiplied by \\(a^2\\). Autocorrelation is also a second-order description, not a complete specification of a general process. Distinct non-Gaussian processes can have identical means and autocovariances while differing in asymmetry, tail behavior, or nonlinear dependence. For jointly Gaussian variables, however, zero covariance implies independence, giving autocorrelation a stronger interpretation."
    },
    {
      "heading": "Characteristic patterns and elementary examples",
      "text": "The simplest reference case is white noise: a sequence with constant mean, constant finite variance, and zero covariance at every nonzero lag. Its normalized autocorrelation is\n\n\\[\\rho(k)= \\begin{cases} 1,&k=0,\\\\ 0,&k\\neq0. \\end{cases}\\]\n\nWhite noise need not consist of independent observations unless independence is imposed separately. Gaussian white noise is independent across time because its finite collections are jointly normal and uncorrelated. This distinction separates the second-order concept of whiteness from stronger notions of randomness.\n\nA first-order autoregressive process provides a basic model of persistence:\n\n\\[X_t=\\phi X_{t-1}+\\varepsilon_t, \\qquad |\\phi|<1,\\]\n\nwhere the innovations are white noise and uncorrelated with earlier values of the process. The stationary solution has\n\n\\[\\gamma(0)=\\frac{\\sigma_\\varepsilon^2}{1-\\phi^2}, \\qquad \\rho(k)=\\phi^{|k|}.\\]\n\nFor \\(0<\\phi<1\\), correlations decay geometrically and remain positive. For \\(-1<\\phi<0\\), their signs alternate. A value of \\(\\phi\\) close to one produces long-lasting dependence, but \\(\\phi=1\\) gives a random walk, which has no constant finite unconditional variance and therefore no stationary autocorrelation function of this form.\n\nA first-order moving-average process instead has finite dependence in its covariance:\n\n\\[X_t=\\varepsilon_t+\\theta\\varepsilon_{t-1}.\\]\n\nIts autocorrelation is\n\n\\[\\rho(1)=\\frac{\\theta}{1+\\theta^2}, \\qquad \\rho(k)=0\\quad\\text{for }|k|>1.\\]\n\nOnly adjacent observations share an innovation. More generally, a moving-average process of order \\(q\\) has zero autocovariance beyond lag \\(q\\). The name describes a weighted combination of innovations, not necessarily a rolling average of observed data.\n\nOscillatory autocorrelation arises when a process contains a preferred frequency. Consider\n\n\\[X_t=A\\cos(\\omega_0t+\\Phi),\\]\n\nwhere \\(\\Phi\\) is uniform on \\([0,2\\pi)\\). Its mean is zero, its covariance is\n\n\\[\\gamma(k)=\\frac{A^2}{2}\\cos(\\omega_0k),\\]\n\nand its autocorrelation is \\(\\cos(\\omega_0k)\\). Randomizing the phase makes this example stationary without removing its periodic dependence. Adding independent white noise increases the variance at lag zero while leaving the nonzero-lag covariance unchanged, reducing the normalized peaks.\n\nA decreasing envelope combined with oscillation often indicates a noisy, damped mode rather than a perfectly periodic component. A slowly decreasing positive tail can reflect long-range dependence, but similar empirical shapes can result from an unremoved trend or a mixture of regimes. Zero correlation also does not rule out dependence: if \\(\\varepsilon_t\\) are independent standard normal variables and \\(X_t=\\varepsilon_t\\varepsilon_{t-1}\\), then \\(X_t\\) is uncorrelated across distinct times, while neighboring squared values are dependent because they share \\(\\varepsilon_t^2\\)."
    },
    {
      "heading": "Spectral representation and the Wiener–Khinchin theorem",
      "text": "Autocorrelation and frequency content are two descriptions of the same second-order structure. For a discrete-time stationary process with absolutely summable autocovariance, define the spectral density by\n\n\\[f(\\omega)=\\frac{1}{2\\pi} \\sum_{k=-\\infty}^{\\infty} \\gamma(k)e^{-ik\\omega}, \\qquad -\\pi\\leq\\omega\\leq\\pi.\\]\n\nThe inverse relation is\n\n\\[\\gamma(k)= \\int_{-\\pi}^{\\pi}e^{ik\\omega}f(\\omega)\\,d\\omega.\\]\n\nThese formulas are a discrete-time form of the Wiener–Khinchin theorem, associated with Norbert Wiener and Aleksandr Khinchin through work published in 1930 and 1934. Different conventions place factors of \\(2\\pi\\) differently or use frequency in cycles rather than angular frequency.\n\nSetting \\(k=0\\) in the inverse relation gives\n\n\\[\\gamma(0)=\\int_{-\\pi}^{\\pi}f(\\omega)\\,d\\omega.\\]\n\nThus total spectral power equals variance for a centered process. Positive semidefiniteness of the covariance corresponds to nonnegativity of the spectral density. A white-noise process has the constant spectrum \\(\\sigma^2/(2\\pi)\\), explaining the analogy with white light, which contains a broad range of frequencies.\n\nNot every stationary process has an ordinary spectral density. The general representation uses a finite nonnegative spectral measure \\(F\\):\n\n\\[\\gamma(k)= \\int_{-\\pi}^{\\pi}e^{ik\\omega}\\,dF(\\omega).\\]\n\nA persistent sinusoidal component contributes point masses to this measure. More complicated singular measures are also possible. The general representation therefore cannot always be replaced by the Fourier transform of an integrable function. It is the spectral measure, rather than a smooth spectrum, that is fundamental.\n\nThe shape of the autocorrelation has a direct spectral interpretation. Slow variation and positive persistence emphasize low frequencies. Alternating correlation emphasizes frequencies near the upper end of the discrete-time range. A narrow spectral peak corresponds to an autocorrelation that oscillates over many lags. For the first-order autoregressive process,\n\n\\[f(\\omega)= \\frac{\\sigma_\\varepsilon^2} {2\\pi\\left(1+\\phi^2-2\\phi\\cos\\omega\\right)}.\\]\n\nAs positive \\(\\phi\\) approaches one, the spectrum becomes increasingly concentrated near zero frequency.\n\nLinear filtering makes the connection especially useful. If a stationary input passes through a stable linear filter with frequency response \\(H(\\omega)\\), then its output spectrum is\n\n\\[f_Y(\\omega)=|H(\\omega)|^2f_X(\\omega).\\]\n\nThe corresponding covariance can be recovered by inverse transformation. Filtering therefore modifies autocorrelation in a predictable way, whether the operation is smoothing, differencing, or selecting a frequency band. Conversely, an autocorrelation alone generally does not reveal a signal’s Fourier phase, so it cannot uniquely reconstruct the original waveform."
    },
    {
      "heading": "Estimation from finite data and numerical computation",
      "text": "For observations \\(x_1,\\ldots,x_n\\), let\n\n\\[\\bar{x}=\\frac1n\\sum_{t=1}^{n}x_t.\\]\n\nA common sample autocovariance estimator is\n\n\\[\\widehat{\\gamma}(k)= \\frac1n\\sum_{t=1}^{n-k} (x_{t+k}-\\bar{x})(x_t-\\bar{x}), \\qquad 0\\leq k<n,\\]\n\nwith sample autocorrelation\n\n\\[\\widehat{\\rho}(k)= \\frac{\\widehat{\\gamma}(k)}{\\widehat{\\gamma}(0)}.\\]\n\nThe graph of these estimates against lag is often called a correlogram. The denominator cancels in the ratio, leaving the lagged cross-product sum divided by the total centered sum of squares.\n\nAnother convention divides the autocovariance sum by \\(n-k\\). If the true mean is known and used for centering, this removes the finite-pair-count bias:\n\n\\[\\operatorname{E}\\!\\left[ \\frac{1}{n-k}\\sum_{t=1}^{n-k} (X_{t+k}-\\mu)(X_t-\\mu) \\right]=\\gamma(k).\\]\n\nWith denominator \\(n\\), the expectation is instead \\((1-k/n)\\gamma(k)\\). Replacing the true mean by the sample mean introduces additional bias, so the \\(n-k\\) convention is not generally exactly unbiased in ordinary applications.\n\nThe denominator \\(n\\) has an important structural advantage. Extend the centered data by zeros outside their observed interval. Their full lagged-product sequence is positive semidefinite, and its Fourier transform is nonnegative. Dividing individual lags by different factors can destroy that property. Consequently, an estimator that improves bias at a particular lag may produce an invalid covariance sequence when all estimated lags are considered together. Estimated autocorrelations, being ratios of random quantities, have their own bias beyond that of either component.\n\nDirect computation through maximum lag \\(K\\) takes \\(O(nK)\\) arithmetic operations. Computing every lag directly takes \\(O(n^2)\\). The fast Fourier transform reduces full autocorrelation computation to \\(O(n\\log n)\\): center the observations, zero-pad them, transform them, multiply the transform by its complex conjugate, and apply the inverse transform. Padding to at least \\(2n-1\\) samples prevents wraparound and yields linear rather than circular correlation. The final scaling depends on the transform convention and desired covariance normalization.\n\nLarge lags use few overlapping pairs and are correspondingly uncertain. Near lag \\(n-1\\), a single product determines the numerator. Missing observations add another complication: using only available pairs gives different lag-specific sample sizes and may break positive semidefiniteness. Irregular sampling is more fundamental still, because a separation of three observation indices need not represent a fixed elapsed time. In such cases, covariance models defined on actual time differences or specialized pair-binning estimators distinguish temporal separation from observation order."
    },
    {
      "heading": "Sampling uncertainty and tests of serial dependence",
      "text": "An estimated autocorrelation fluctuates even when the population autocorrelation is zero. For independent, identically distributed observations with suitable finite moments, the sample autocorrelation at a fixed nonzero lag is asymptotically normal with variance approximately \\(1/n\\). This motivates the familiar reference bands\n\n\\[\\pm\\frac{1.96}{\\sqrt n}.\\]\n\nThey are approximate pointwise 95% bands under a white-noise null model, not universal confidence intervals for arbitrary processes. They also assume a nondegenerate variance and an asymptotic regime in which the chosen lag remains fixed as sample size grows.\n\nExamining many lags changes the interpretation. Even with independent observations, some sample correlations can fall outside pointwise bands by chance. The lag estimates are generally dependent, so the graph is not a collection of independent tests. This is a multiple-testing issue: a single unusual bar has a different evidential meaning from a preselected test at one lag or a sustained, structured pattern across many lags.\n\nWhen the process is already correlated, the white-noise variance approximation is inappropriate. Bartlett-type approximations express the covariance of sample autocorrelations through the population correlation sequence. For a moving-average process of order \\(q\\), at lags \\(k>q\\) a standard asymptotic expression is\n\n\\[\\operatorname{Var}[\\widehat{\\rho}(k)] \\approx \\frac1n\\left(1+2\\sum_{j=1}^{q}\\rho(j)^2\\right).\\]\n\nThe increased variance reflects information shared across observations. Such formulas require assumptions controlling moments and dependence; they are not automatically valid for long-memory processes or data with undefined second moments.\n\nJoint tests aggregate several lag correlations. The Box–Pierce statistic, introduced by George Box and David Pierce in 1970, is\n\n\\[Q_{\\mathrm{BP}}= n\\sum_{k=1}^{h}\\widehat{\\rho}(k)^2.\\]\n\nThe Ljung–Box modification, introduced by Greta Ljung and Box in 1978, is\n\n\\[Q_{\\mathrm{LB}}= n(n+2)\\sum_{k=1}^{h} \\frac{\\widehat{\\rho}(k)^2}{n-k}.\\]\n\nUnder an appropriate white-noise null and regularity conditions, these are compared with a chi-squared distribution. For residuals from a fitted autoregressive moving-average model, a common asymptotic degrees-of-freedom adjustment is \\(h-p-q\\), where \\(p\\) and \\(q\\) are the fitted autoregressive and moving-average orders and \\(h\\) exceeds their sum.\n\nThese portmanteau tests address collections of linear correlations. Failure to reject does not establish independence, Gaussianity, constant conditional variance, or general model correctness. A process can have uncorrelated observations but correlated squares or absolute values. Residual diagnostics therefore often examine transformations as well as raw residuals. Parameter estimation, conditional heteroskedasticity, and data-dependent lag selection can alter reference distributions, making the precise null model central to hypothesis testing."
    },
    {
      "heading": "Linear prediction, autoregressive models, and partial autocorrelation",
      "text": "Autocovariance determines optimal linear prediction whenever the relevant second moments exist. Suppose a centered stationary process is predicted from its last \\(p\\) observations:\n\n\\[\\widehat X_t= \\sum_{j=1}^{p}a_jX_{t-j}.\\]\n\nMinimizing mean squared error gives the normal equations\n\n\\[\\sum_{j=1}^{p}a_j\\gamma(i-j)=\\gamma(i), \\qquad i=1,\\ldots,p.\\]\n\nThese follow by making the prediction error uncorrelated with every predictor. Geometrically, prediction is an orthogonal projection onto the span of past observations in a space of square-integrable random variables.\n\nFor an autoregressive process of order \\(p\\),\n\n\\[X_t=\\phi_1X_{t-1}+\\cdots+\\phi_pX_{t-p}+\\varepsilon_t,\\]\n\nthe corresponding relations are the Yule–Walker equations:\n\n\\[\\gamma(k)=\\sum_{j=1}^{p}\\phi_j\\gamma(k-j), \\qquad k\\geq1.\\]\n\nTheir name reflects foundational work by G. Udny Yule in 1927 and Gilbert Walker in 1931. Together with\n\n\\[\\gamma(0)= \\sum_{j=1}^{p}\\phi_j\\gamma(j)+\\sigma_\\varepsilon^2,\\]\n\nthey connect model parameters, innovation variance, and autocovariance. Replacing population covariances by estimates produces a standard parameter-estimation method.\n\nThe partial autocorrelation function separates a direct linear association at lag \\(k\\) from association mediated through intervening observations. For a real stationary process, it is the correlation between the residuals obtained by linearly regressing \\(X_t\\) and \\(X_{t-k}\\) on \\(X_{t-1},\\ldots,X_{t-k+1}\\). Equivalently, under nonsingularity, it is the final coefficient in the order-\\(k\\) linear prediction equations.\n\nThis distinction supports classical model identification. A stationary autoregressive process of order \\(p\\) has partial autocorrelations equal to zero beyond \\(p\\), while its ordinary autocorrelation usually tails off. A moving-average process of order \\(q\\) has ordinary autocorrelation equal to zero beyond \\(q\\), while its partial autocorrelation generally tails off. The population patterns are exact, but finite-sample plots only approximate them, sometimes poorly when parameters are near cancellation or near a stationarity boundary.\n\nA mixed autoregressive moving-average model can produce both decaying and oscillatory correlation patterns. For a causal stationary autoregressive representation, the roots of\n\n\\[1-\\phi_1z-\\cdots-\\phi_pz^p=0\\]\n\nlie outside the unit circle. Their locations determine the rates and frequencies of covariance decay. Efficient recursions exploit the Toeplitz structure of the prediction equations, connecting autocorrelation analysis with numerical linear algebra. The resulting predictor is optimal among linear predictors; it equals the full conditional expectation only under additional distributional structure, such as joint Gaussianity."
    },
    {
      "heading": "Nonstationarity, long memory, and interpretation limits",
      "text": "A large empirical autocorrelation does not by itself identify a stable stochastic mechanism. A deterministic trend can make nearby observations similar simply because both lie on the same slowly changing trajectory. Computing the sample ACF after subtracting only a single overall mean can then produce a long positive tail, even when the deviations around the trend are independent. Seasonal means, level shifts, and changes in measurement practice can create similarly misleading patterns.\n\nDifferencing changes the dependence structure rather than merely removing a graphical feature. If\n\n\\[Y_t=X_t-X_{t-1},\\]\n\nand \\(X_t\\) is stationary, then\n\n\\[\\gamma_Y(k)= 2\\gamma_X(k)-\\gamma_X(k-1)-\\gamma_X(k+1).\\]\n\nIn frequency terms, differencing multiplies the spectrum by\n\n\\[|1-e^{-i\\omega}|^2= 4\\sin^2(\\omega/2),\\]\n\nsuppressing low frequencies. Differencing can transform a random walk into its stationary increments, but unnecessary differencing can introduce negative short-lag correlation. For instance, differenced independent noise has lag-one correlation \\(-1/2\\).\n\nTrue long-range dependence is distinct from a trend. One common form has\n\n\\[\\rho(k)\\sim Ck^{-\\alpha}, \\qquad 0<\\alpha<1,\\quad C>0.\\]\n\nThe positive autocorrelations are not summable. Equivalently, under suitable conditions, the spectral density diverges near zero frequency. Such a long-memory process can have a constant mean and variance while retaining dependence over widely separated times. The variance of its sample mean often decreases more slowly than the usual \\(1/n\\) rate, so ordinary short-memory uncertainty formulas no longer apply.\n\nTemporal aggregation also changes autocorrelation. Monthly averages of daily observations are filtered and subsampled versions of the original process; their dependence is not simply the daily autocorrelation evaluated every thirty lags. Sampling itself can cause aliasing, in which distinct continuous-time frequencies become indistinguishable on a discrete observation grid. Therefore, an oscillatory correlogram must be interpreted relative to the measurement interval and the acquisition procedure.\n\nAutocorrelation is symmetric in lag for real stationary processes and consequently does not encode a direction of influence. It cannot, by itself, establish causation, distinguish an external driver from internal dynamics, or determine whether an apparent cycle is generated by feedback. Moreover, processes with infinite variance, including many heavy-tailed models, lack the conventional covariance-based ACF altogether. Rank-based dependence measures, characteristic-function methods, and other alternatives describe different aspects of temporal dependence, rather than repairing an undefined second moment."
    },
    {
      "heading": "Deterministic signals, detection, and physical measurements",
      "text": "For a finite-energy signal, autocorrelation has a direct geometric interpretation. The value\n\n\\[R_x(\\tau)= \\int x(t+\\tau)\\overline{x(t)}\\,dt\\]\n\nis an inner product between the signal and a translated copy. At zero lag it equals the signal energy,\n\n\\[R_x(0)=\\int |x(t)|^2\\,dt,\\]\n\nand the Cauchy–Schwarz inequality gives\n\n\\[|R_x(\\tau)|\\leq R_x(0).\\]\n\nA peak at a nonzero lag indicates that the signal resembles a shifted version of itself. Repeated peaks can expose a period even when the waveform is complicated or contaminated by noise.\n\nAutocorrelation can also be written as a convolution:\n\n\\[R_x(\\tau)= (x*\\widetilde{x})(\\tau), \\qquad \\widetilde{x}(t)=\\overline{x(-t)}.\\]\n\nIts Fourier transform is \\(|X(\\omega)|^2\\), where \\(X\\) is the transform of \\(x\\). This is the deterministic energy-signal counterpart of the stationary-process spectral relation. Because the transform phase disappears, different signals can share the same autocorrelation. A time shift, for example, leaves the autocorrelation unchanged. The associated reconstruction ambiguity is central to phase retrieval.\n\nIn radar and sonar, the correlation of a received waveform with a known transmitted waveform is technically cross-correlation. When the received echo is an attenuated, delayed copy, its delay response is governed by the transmitted waveform’s autocorrelation. A narrow central peak improves delay discrimination, while large sidelobes can obscure weaker nearby echoes. This is one reason coded pulses and carefully designed sequences are used. For radar propagation in a uniform medium, a round-trip delay \\(\\tau\\) corresponds to range \\(v\\tau/2\\), where \\(v\\) is the propagation speed.\n\nIn speech and music analysis, a quasi-periodic waveform produces autocorrelation peaks near integer multiples of its fundamental period. If the sampling rate is \\(f_s\\) and a relevant peak occurs at lag \\(k_0\\), a candidate fundamental frequency is \\(f_s/k_0\\). Harmonics, changes in pitch, finite windows, and noise can create competing peaks, so the largest nonzero-lag peak does not always identify the perceived fundamental.\n\nPhysical correlation functions often involve a measured observable rather than the raw position or field. For an isotropic equilibrium fluid in three dimensions, the Green–Kubo relation for diffusion is\n\n\\[D=\\frac13\\int_0^\\infty \\left\\langle \\mathbf v(t)\\cdot\\mathbf v(0) \\right\\rangle\\,dt,\\]\n\nwhen the equilibrium and long-time limits required by the relation exist. The velocity autocorrelation therefore links microscopic dynamics to a macroscopic transport coefficient. Related equilibrium time-correlation formulas connect stress fluctuations with viscosity and heat-current fluctuations with thermal conductivity; the observable, normalization, and physical assumptions differ in each case."
    },
    {
      "heading": "Correlated simulations, spatial fields, and multivariate extensions",
      "text": "Autocorrelation directly affects the uncertainty of an average. For \\(n\\) observations from a stationary real process,\n\n\\[\\operatorname{Var}(\\bar X_n)= \\frac{\\gamma(0)}{n} \\left[ 1+2\\sum_{k=1}^{n-1} \\left(1-\\frac{k}{n}\\right)\\rho(k) \\right].\\]\n\nThis identity follows by summing all entries of the covariance matrix. It shows why \\(n\\) correlated measurements do not generally carry the same information about the mean as \\(n\\) independent measurements. Positive correlations increase the variance; sufficiently structured negative correlations can reduce it.\n\nIf the correlations are absolutely summable, define, under one common convention,\n\n\\[\\tau_{\\mathrm{int}}= 1+2\\sum_{k=1}^{\\infty}\\rho(k).\\]\n\nThen, for large \\(n\\),\n\n\\[\\operatorname{Var}(\\bar X_n) \\approx \\frac{\\gamma(0)\\tau_{\\mathrm{int}}}{n}.\\]\n\nThe quantity \\(\\tau_{\\mathrm{int}}\\) is the integrated autocorrelation time. Another convention assigns half this value, so numerical comparisons require checking the definition. For a first-order autoregressive process,\n\n\\[\\tau_{\\mathrm{int}}=\\frac{1+\\phi}{1-\\phi}.\\]\n\nUnder the corresponding asymptotic mean-variance interpretation, an effective sample size is \\(n/\\tau_{\\mathrm{int}}\\).\n\nThese formulas are especially important in Markov chain Monte Carlo. Successive simulated states are intentionally dependent, and each scalar observable can have a different autocorrelation time. A rapidly mixing coordinate does not establish that every parameter or derived quantity is well explored. Estimating an infinite sum from a finite chain requires truncation or smoothing because large-lag estimates are noisy. Autocorrelation calculations also presuppose an appropriate stationary regime; they cannot reveal an unvisited region of the target distribution simply from fluctuations inside the visited region.\n\nFor a vector-valued process \\(X_t\\), autocovariance becomes matrix-valued:\n\n\\[\\Gamma(k)= \\operatorname{E}\\big[ (X_{t+k}-\\mu)(X_t-\\mu)^\\ast \\big].\\]\n\nIts diagonal entries are component autocovariances, and its off-diagonal entries describe cross-dependence at different lags. Hermitian symmetry becomes \\(\\Gamma(-k)=\\Gamma(k)^\\ast\\), and the spectral counterpart is a positive-semidefinite matrix at each frequency. Unlike a scalar real ACF, an individual cross-covariance between two different components need not be even in lag.\n\nSpatial random fields replace temporal lag by a displacement vector \\(h\\). A stationary covariance \\(C(h)\\) may depend on direction as well as distance; under isotropy it depends only on \\(\\|h\\|\\). For a real stationary field, the semivariogram is\n\n\\[v(h)=\\frac12 \\operatorname{E}\\big[(X(s+h)-X(s))^2\\big] =C(0)-C(h).\\]\n\nThis relation links autocovariance with spatial interpolation and geostatistical modeling. Validity still requires positive semidefiniteness over every finite set of locations, and a radial function admissible in one spatial dimension need not be admissible in every higher dimension."
    }
  ],
  "links": [
    "time-series",
    "stochastic-process",
    "expected-value",
    "variance",
    "joint-probability-distribution",
    "ergodicity",
    "cauchy-schwarz-inequality",
    "gaussian-process",
    "statistical-independence",
    "autoregressive-model",
    "random-walk",
    "moving-average-model",
    "wiener-khinchin-theorem",
    "norbert-wiener",
    "fourier-transform",
    "sample-mean",
    "estimator-bias",
    "fast-fourier-transform",
    "confidence-interval",
    "multiple-testing",
    "chi-squared-distribution",
    "statistical-hypothesis-testing",
    "mean-squared-error",
    "orthogonal-projection",
    "numerical-linear-algebra",
    "aliasing",
    "causation",
    "inner-product",
    "convolution",
    "markov-chain-monte-carlo"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}