# Stationary process

> Mediated Wiki article. Canonical URL: https://mediated.wiki/source/Stationary_process
> Markdown URL: https://mediated.wiki/source/Stationary_process.md
> Source: https://en.wikipedia.org/wiki/Stationary_process
> Source revision: 1342754986
> License: Creative Commons Attribution-ShareAlike 4.0 International (https://creativecommons.org/licenses/by-sa/4.0/)

Type of stochastic process

In [mathematics](/source/Mathematics) and [statistics](/source/Statistics), a **stationary process** (also called a **strict/strictly stationary process** or **strong/strongly stationary process**) is a [stochastic process](/source/Stochastic_process) whose statistical properties, such as [mean](/source/Mean) and [variance](/source/Variance), do not change over time. More formally, the [joint probability distribution](/source/Joint_probability_distribution) of the process remains the same when shifted in time. This implies that the process is statistically consistent across different time periods. Because many statistical procedures in [time series analysis](/source/Time_series_analysis) assume stationarity, non-stationary data are frequently transformed to achieve stationarity before analysis.

A common cause of non-stationarity is a trend in the mean, which can be due to either a [unit root](/source/Unit_root) or a deterministic trend. In the case of a unit root, stochastic shocks have permanent effects, and the process is not [mean-reverting](/source/Mean-reverting_process). With a deterministic trend, the process is called [trend-stationary](/source/Trend-stationary_process), and shocks have only transitory effects, with the variable tending towards a deterministically evolving mean. A trend-stationary process is not strictly stationary but can be made stationary by removing the trend. Similarly, processes with unit roots can be made stationary through [differencing](/source/Differencing).

Another type of non-stationary process, distinct from those with trends, is a [cyclostationary process](/source/Cyclostationary_process), which exhibits cyclical variations over time.

Strict stationarity, as defined above, can be too restrictive for many applications. Therefore, other forms of stationarity, such as **wide-sense stationarity** or **⁠ N {\displaystyle N} ⁠th-order stationarity**, are often used. The definitions for different kinds of stationarity are not consistent among different authors (see [Other terminology](#Other_terminology)).

## Strict-sense stationarity

### Definition

Formally, let { X t } {\displaystyle \left\{X_{t}\right\}} be a [stochastic process](/source/Stochastic_process) and let F X ( x t 1 + τ , … , x t n + τ ) {\displaystyle F_{X}(x_{t_{1}+\tau },\ldots ,x_{t_{n}+\tau })} represent the [cumulative distribution function](/source/Cumulative_distribution_function) of the [unconditional](/source/Marginal_distribution) (i.e., with no reference to any particular starting value) [joint distribution](/source/Joint_distribution) of { X t } {\displaystyle \left\{X_{t}\right\}} at times ⁠ t 1 + τ , … , t n + τ {\displaystyle t_{1}+\tau ,\ldots ,t_{n}+\tau } ⁠. Then, { X t } {\displaystyle \left\{X_{t}\right\}} is said to be **strictly stationary**, **strongly stationary** or **strict-sense stationary** if[1]: 155

F X ( x t 1 + τ , … , x t n + τ ) = F X ( x t 1 , … , x t n ) for all τ , t 1 , … , t n ∈ R and for all n ∈ N > 0 {\displaystyle F_{X}(x_{t_{1}+\tau },\ldots ,x_{t_{n}+\tau })=F_{X}(x_{t_{1}},\ldots ,x_{t_{n}})\quad {\text{for all }}\tau ,t_{1},\ldots ,t_{n}\in \mathbb {R} {\text{ and for all }}n\in \mathbb {N} _{>0}} Eq.1

Since τ {\displaystyle \tau } does not affect ⁠ F X ( ⋅ ) {\displaystyle F_{X}(\cdot )} ⁠, F X {\displaystyle F_{X}} is independent of time.

### Examples

Two simulated time series processes, one stationary and the other non-stationary, are shown above. The [augmented Dickey–Fuller](/source/Augmented_Dickey-Fuller_test) (ADF) [test statistic](/source/Test_statistic) is reported for each process; non-stationarity cannot be rejected for the second process at a 5% [significance level](/source/Significance_level).

[White noise](/source/White_noise) is the simplest example of a stationary process.

An example of a [discrete-time](/source/Discrete-time_stochastic_process) stationary process where the sample space is also discrete (so that the random variable may take one of ⁠ N {\displaystyle N} ⁠ possible values) is a [Bernoulli scheme](/source/Bernoulli_scheme). Other examples of a discrete-time stationary process with continuous sample space include some [autoregressive](/source/Autoregressive) and [moving average](/source/Moving_average_model) processes that are both subsets of the [autoregressive moving average model](/source/Autoregressive_moving_average_model). Models with a non-trivial autoregressive component may be either stationary or non-stationary, depending on the parameter values, and important non-stationary special cases are where [unit roots](/source/Unit_root) exist in the model.

#### Example 1

Let Y {\displaystyle Y} be any scalar [random variable](/source/Random_variable), and define a time-series { X t } {\displaystyle \left\{X_{t}\right\}} by

- X t = Y for all t . {\displaystyle X_{t}=Y\qquad {\text{ for all }}t.}

Then { X t } {\displaystyle \left\{X_{t}\right\}} is a stationary time series, for which realisations consist of a series of constant values, with a different constant value for each realisation. A [law of large numbers](/source/Law_of_large_numbers) does not apply on this case, as the limiting value of an average from a single realisation takes the random value determined by ⁠ Y {\displaystyle Y} ⁠, rather than taking the [expected value](/source/Expected_value) of ⁠ Y {\displaystyle Y} ⁠.

The time average of X t {\displaystyle X_{t}} does not converge since the process is not [ergodic](/source/Ergodic_process).

#### Example 2

As a further example of a stationary process for which any single realisation has an apparently noise-free structure, let Y {\displaystyle Y} have a [uniform distribution](/source/Uniform_distribution_(continuous)) on [ 0 , 2 π ] {\displaystyle [0,2\pi ]} and define the time series { X t } {\displaystyle \left\{X_{t}\right\}} by

- X t = cos ⁡ ( t + Y ) for t ∈ R . {\displaystyle X_{t}=\cos(t+Y)\quad {\text{ for }}t\in \mathbb {R} .}

Then { X t } {\displaystyle \left\{X_{t}\right\}} is strictly stationary since (⁠ ( t + Y ) {\displaystyle (t+Y)} ⁠ modulo ⁠ 2 π {\displaystyle 2\pi } ⁠) follows the same uniform distribution as Y {\displaystyle Y} for any ⁠ t {\displaystyle t} ⁠.

#### Example 3

Keep in mind that a [weakly white noise](/source/White_noise) is not necessarily strictly stationary. Let ω {\displaystyle \omega } be a random variable uniformly distributed in the interval ( 0 , 2 π ) {\displaystyle (0,2\pi )} and define the time series { z t } {\displaystyle \left\{z_{t}\right\}} by

- z t = cos ⁡ ( t ω ) ( t = 1 , 2 , . . . ) {\displaystyle z_{t}=\cos(t\omega )\quad (t=1,2,...)}

Then

- E ( z t ) = 1 2 π ∫ 0 2 π cos ⁡ ( t ω ) d ω = 0 , Var ⁡ ( z t ) = 1 2 π ∫ 0 2 π cos 2 ⁡ ( t ω ) d ω = 1 / 2 , Cov ⁡ ( z t , z j ) = 1 2 π ∫ 0 2 π cos ⁡ ( t ω ) cos ⁡ ( j ω ) d ω = 0 ∀ t ≠ j . {\displaystyle {\begin{aligned}\mathbb {E} (z_{t})&={\frac {1}{2\pi }}\int _{0}^{2\pi }\cos(t\omega )\,d\omega =0,\\\operatorname {Var} (z_{t})&={\frac {1}{2\pi }}\int _{0}^{2\pi }\cos ^{2}(t\omega )\,d\omega =1/2,\\\operatorname {Cov} (z_{t},z_{j})&={\frac {1}{2\pi }}\int _{0}^{2\pi }\cos(t\omega )\cos(j\omega )\,d\omega =0\quad \forall t\neq j.\end{aligned}}}

So { z t } {\displaystyle \{z_{t}\}} is a white noise in the weak sense (the mean and cross-covariances are zero, and the variances are all the same), however it is not strictly stationary.

## *N*th-order stationarity

In **[Eq.1](#math_Eq.1)**, the distribution of n {\displaystyle n} samples of the stochastic process must be equal to the distribution of the samples shifted in time *for all* ⁠ n {\displaystyle n} ⁠. ⁠ N {\displaystyle N} ⁠th-order stationarity is a weaker form of stationarity where this is only requested for all n {\displaystyle n} up to a certain order ⁠ N {\displaystyle N} ⁠. A random process { X t } {\displaystyle \left\{X_{t}\right\}} is said to be **⁠ N {\displaystyle N} ⁠th-order stationary** if:[1]: 152

F X ( x t 1 + τ , … , x t n + τ ) = F X ( x t 1 , … , x t n ) for all τ , t 1 , … , t n ∈ R and for all n ∈ { 1 , … , N } {\displaystyle F_{X}(x_{t_{1}+\tau },\ldots ,x_{t_{n}+\tau })=F_{X}(x_{t_{1}},\ldots ,x_{t_{n}})\quad {\text{for all }}\tau ,t_{1},\ldots ,t_{n}\in \mathbb {R} {\text{ and for all }}n\in \{1,\ldots ,N\}} Eq.2

## Weak or wide-sense stationarity

### Definition

A weaker form of stationarity commonly employed in [signal processing](/source/Signal_processing) is known as **weak-sense stationarity**, **wide-sense stationarity (WSS)**, or **covariance stationarity**. WSS random processes only require that 1st [moment](/source/Moment_(mathematics)) (i.e. the mean) and [autocovariance](/source/Autocovariance) do not vary with respect to time and that the 2nd moment is finite for all times. Any strictly stationary process that has a finite [mean](/source/Mean) and [covariance](/source/Covariance) is also WSS.[2]: 299

So, a [continuous time](/source/Continuous_time) [random process](/source/Random_process) { X t } {\displaystyle \left\{X_{t}\right\}} that is WSS has the following restrictions on its mean function m X ( t ) ≜ E ⁡ [ X t ] {\displaystyle m_{X}(t)\triangleq \operatorname {E} [X_{t}]} and [autocovariance](/source/Autocovariance) function ⁠ K X X ( t 1 , t 2 ) ≜ E ⁡ [ ( X t 1 − m X ( t 1 ) ) ( X t 2 − m X ( t 2 ) ) ] {\displaystyle K_{XX}(t_{1},t_{2})\triangleq \operatorname {E} [(X_{t_{1}}-m_{X}(t_{1}))(X_{t_{2}}-m_{X}(t_{2}))]} ⁠:

m X ( t ) = m X ( t + τ ) for all τ , t ∈ R K X X ( t 1 , t 2 ) = K X X ( t 1 − t 2 , 0 ) for all t 1 , t 2 ∈ R E ⁡ [ | X t | 2 ] < ∞ for all t ∈ R {\displaystyle {\begin{aligned}&m_{X}(t)=m_{X}(t+\tau )&&{\text{for all }}\tau ,t\in \mathbb {R} \\&K_{XX}(t_{1},t_{2})=K_{XX}(t_{1}-t_{2},0)&&{\text{for all }}t_{1},t_{2}\in \mathbb {R} \\&\operatorname {E} [|X_{t}|^{2}]<\infty &&{\text{for all }}t\in \mathbb {R} \end{aligned}}} Eq.3

The first property implies that the mean function m X ( t ) {\displaystyle m_{X}(t)} must be constant. The second property implies that the autocovariance function depends only on the *difference* between t 1 {\displaystyle t_{1}} and t 2 {\displaystyle t_{2}} and only needs to be indexed by one variable rather than two variables.[1]: 159 Thus, instead of writing,

- K X X ( t 1 − t 2 , 0 ) {\displaystyle \,\!K_{XX}(t_{1}-t_{2},0)\,}

the notation is often abbreviated by the substitution ⁠ τ = t 1 − t 2 {\displaystyle \tau =t_{1}-t_{2}} ⁠:

- K X X ( τ ) ≜ K X X ( t 1 − t 2 , 0 ) {\displaystyle K_{XX}(\tau )\triangleq K_{XX}(t_{1}-t_{2},0)}

This also implies that the [autocorrelation](/source/Autocorrelation) depends only on ⁠ τ = t 1 − t 2 {\displaystyle \tau =t_{1}-t_{2}} ⁠, that is

- R X ( t 1 , t 2 ) = R X ( t 1 − t 2 , 0 ) ≜ R X ( τ ) . {\displaystyle R_{X}(t_{1},t_{2})=R_{X}(t_{1}-t_{2},0)\triangleq R_{X}(\tau ).}

The third property says that the second moments must be finite for any time ⁠ t {\displaystyle t} ⁠.

### Motivation

The main advantage of wide-sense stationarity is that it places the time-series in the context of [Hilbert spaces](/source/Hilbert_space). Let ⁠ H {\displaystyle H} ⁠ be the Hilbert space generated by ⁠ { x ( t ) } {\displaystyle \{x(t)\}} ⁠ (that is, the closure of the set of all linear combinations of these random variables in the Hilbert space of all square-integrable random variables on the given probability space). By the positive definiteness of the autocovariance function, it follows from [Bochner's theorem](/source/Bochner's_theorem) that there exists a positive measure μ {\displaystyle \mu } on the real line such that ⁠ H {\displaystyle H} ⁠ is isomorphic to the Hilbert subspace of ⁠ L 2 ( μ ) {\displaystyle L^{2}(\mu )} ⁠ generated by ⁠ { e − 2 π i ξ ⋅ t } {\displaystyle \{e^{-2\pi i\xi \cdot t}\}} ⁠. This then gives the following Fourier-type decomposition for a continuous time stationary stochastic process: there exists a stochastic process ω ξ {\displaystyle \omega _{\xi }} with [orthogonal increments](https://en.wikipedia.org/w/index.php?title=Orthogonal_increments&action=edit&redlink=1) such that, for all ⁠ t {\displaystyle t} ⁠:

- X t = ∫ e − 2 π i λ ⋅ t d ω λ , {\displaystyle X_{t}=\int e^{-2\pi i\lambda \cdot t}\,d\omega _{\lambda },}

where the integral on the right-hand side is interpreted in a suitable (Riemann) sense. The same result holds for a discrete-time stationary process, with the spectral measure now defined on the unit circle.

When processing WSS random signals with [linear](/source/Linear), [time-invariant](/source/Time-invariant) ([LTI](/source/LTI_system_theory)) [filters](/source/Filter_(signal_processing)), it is helpful to think of the correlation function as a [linear operator](/source/Linear_operator). Since it is a [circulant](/source/Circulant_matrix) operator (depends only on the difference between the two arguments), its eigenfunctions are the [Fourier](/source/Fourier_series) complex exponentials. Additionally, since the [eigenfunctions](/source/Eigenfunction) of LTI operators are also [complex exponentials](/source/Exponential_function), LTI processing of WSS random signals is highly tractable—all computations can be performed in the [frequency domain](/source/Frequency_domain). Thus, the WSS assumption is widely employed in signal processing [algorithms](/source/Algorithm).

### Definition for complex stochastic process

In the case where { X t } {\displaystyle \left\{X_{t}\right\}} is a complex stochastic process the [autocovariance](/source/Autocovariance) function is defined as K X X ( t 1 , t 2 ) = E ⁡ [ ( X t 1 − m X ( t 1 ) ) ( X t 2 − m X ( t 2 ) ) ¯ ] {\displaystyle K_{XX}(t_{1},t_{2})=\operatorname {E} [(X_{t_{1}}-m_{X}(t_{1})){\overline {(X_{t_{2}}-m_{X}(t_{2}))}}]} and, in addition to the requirements in **[Eq.3](#math_Eq.3)**, it is required that the pseudo-autocovariance function J X X ( t 1 , t 2 ) = E ⁡ [ ( X t 1 − m X ( t 1 ) ) ( X t 2 − m X ( t 2 ) ) ] {\displaystyle J_{XX}(t_{1},t_{2})=\operatorname {E} [(X_{t_{1}}-m_{X}(t_{1}))(X_{t_{2}}-m_{X}(t_{2}))]} depends only on the time lag. In formulas, { X t } {\displaystyle \left\{X_{t}\right\}} is WSS, if

m X ( t ) = m X ( t + τ ) for all τ , t ∈ R K X X ( t 1 , t 2 ) = K X X ( t 1 − t 2 , 0 ) for all t 1 , t 2 ∈ R J X X ( t 1 , t 2 ) = J X X ( t 1 − t 2 , 0 ) for all t 1 , t 2 ∈ R E ⁡ [ | X ( t ) | 2 ] < ∞ for all t ∈ R {\displaystyle {\begin{aligned}&m_{X}(t)=m_{X}(t+\tau )&&{\text{for all }}\tau ,t\in \mathbb {R} \\&K_{XX}(t_{1},t_{2})=K_{XX}(t_{1}-t_{2},0)&&{\text{for all }}t_{1},t_{2}\in \mathbb {R} \\&J_{XX}(t_{1},t_{2})=J_{XX}(t_{1}-t_{2},0)&&{\text{for all }}t_{1},t_{2}\in \mathbb {R} \\&\operatorname {E} [|X(t)|^{2}]<\infty &&{\text{for all }}t\in \mathbb {R} \end{aligned}}} Eq.4

## Joint stationarity

The concept of stationarity may be extended to two stochastic processes.

### Joint strict-sense stationarity

Two stochastic processes { X t } {\displaystyle \left\{X_{t}\right\}} and { Y t } {\displaystyle \left\{Y_{t}\right\}} are called **jointly strict-sense stationary** if their joint cumulative distribution F X Y ( x t 1 , … , x t m , y t 1 ′ , … , y t n ′ ) {\displaystyle F_{XY}(x_{t_{1}},\ldots ,x_{t_{m}},y_{t_{1}^{'}},\ldots ,y_{t_{n}^{'}})} remains unchanged under time shifts, i.e. if

F X Y ( x t 1 , … , x t m , y t 1 ′ , … , y t n ′ ) = F X Y ( x t 1 + τ , … , x t m + τ , y t 1 ′ + τ , … , y t n ′ + τ ) for all τ , t 1 , … , t m , t 1 ′ , … , t n ′ ∈ R and for all m , n ∈ N {\displaystyle F_{XY}(x_{t_{1}},\ldots ,x_{t_{m}},y_{t_{1}^{'}},\ldots ,y_{t_{n}^{'}})=F_{XY}(x_{t_{1}+\tau },\ldots ,x_{t_{m}+\tau },y_{t_{1}^{'}+\tau },\ldots ,y_{t_{n}^{'}+\tau })\quad {\text{for all }}\tau ,t_{1},\ldots ,t_{m},t_{1}^{'},\ldots ,t_{n}^{'}\in \mathbb {R} {\text{ and for all }}m,n\in \mathbb {N} } Eq.5

### Joint (*M* + *N*)th-order stationarity

Two random processes { X t } {\displaystyle \left\{X_{t}\right\}} and { Y t } {\displaystyle \left\{Y_{t}\right\}} is said to be **jointly (⁠ M + N {\displaystyle M+N} ⁠)th-order stationary** if:[1]: 159

F X Y ( x t 1 , … , x t m , y t 1 ′ , … , y t n ′ ) = F X Y ( x t 1 + τ , … , x t m + τ , y t 1 ′ + τ , … , y t n ′ + τ ) for all τ , t 1 , … , t m , t 1 ′ , … , t n ′ ∈ R and for all m ∈ { 1 , … , M } , n ∈ { 1 , … , N } {\displaystyle F_{XY}(x_{t_{1}},\ldots ,x_{t_{m}},y_{t_{1}^{'}},\ldots ,y_{t_{n}^{'}})=F_{XY}(x_{t_{1}+\tau },\ldots ,x_{t_{m}+\tau },y_{t_{1}^{'}+\tau },\ldots ,y_{t_{n}^{'}+\tau })\quad {\text{for all }}\tau ,t_{1},\ldots ,t_{m},t_{1}^{'},\ldots ,t_{n}^{'}\in \mathbb {R} {\text{ and for all }}m\in \{1,\ldots ,M\},n\in \{1,\ldots ,N\}} Eq.6

### Joint weak or wide-sense stationarity

Two stochastic processes { X t } {\displaystyle \left\{X_{t}\right\}} and { Y t } {\displaystyle \left\{Y_{t}\right\}} are called **jointly wide-sense stationary** if they are both wide-sense stationary and their cross-covariance function K X Y ( t 1 , t 2 ) = E ⁡ [ ( X t 1 − m X ( t 1 ) ) ( Y t 2 − m Y ( t 2 ) ) ] {\displaystyle K_{XY}(t_{1},t_{2})=\operatorname {E} [(X_{t_{1}}-m_{X}(t_{1}))(Y_{t_{2}}-m_{Y}(t_{2}))]} depends only on the time difference ⁠ τ = t 1 − t 2 {\displaystyle \tau =t_{1}-t_{2}} ⁠. This may be summarized as follows:

m X ( t ) = m X ( t + τ ) for all τ , t ∈ R m Y ( t ) = m Y ( t + τ ) for all τ , t ∈ R K X X ( t 1 , t 2 ) = K X X ( t 1 − t 2 , 0 ) for all t 1 , t 2 ∈ R K Y Y ( t 1 , t 2 ) = K Y Y ( t 1 − t 2 , 0 ) for all t 1 , t 2 ∈ R K X Y ( t 1 , t 2 ) = K X Y ( t 1 − t 2 , 0 ) for all t 1 , t 2 ∈ R {\displaystyle {\begin{aligned}&m_{X}(t)=m_{X}(t+\tau )&&{\text{for all }}\tau ,t\in \mathbb {R} \\&m_{Y}(t)=m_{Y}(t+\tau )&&{\text{for all }}\tau ,t\in \mathbb {R} \\&K_{XX}(t_{1},t_{2})=K_{XX}(t_{1}-t_{2},0)&&{\text{for all }}t_{1},t_{2}\in \mathbb {R} \\&K_{YY}(t_{1},t_{2})=K_{YY}(t_{1}-t_{2},0)&&{\text{for all }}t_{1},t_{2}\in \mathbb {R} \\&K_{XY}(t_{1},t_{2})=K_{XY}(t_{1}-t_{2},0)&&{\text{for all }}t_{1},t_{2}\in \mathbb {R} \end{aligned}}} Eq.7

## Relation between types of stationarity

- If a stochastic process is ⁠ N {\displaystyle N} ⁠th-order stationary, then it is also ⁠ M {\displaystyle M} ⁠th-order stationary for all ⁠ M ≤ N {\displaystyle M\leq N} ⁠.

- If a stochastic process is second order stationary (⁠ N = 2 {\displaystyle N=2} ⁠) and has finite second moments, then it is also wide-sense stationary.[1]: 159

- If a stochastic process is wide-sense stationary, it is not necessarily second-order stationary.[1]: 159

- If a stochastic process is strict-sense stationary and has finite second moments, it is wide-sense stationary.[2]: 299

- If two stochastic processes are jointly (⁠ M + N {\displaystyle M+N} ⁠)th-order stationary, this does not guarantee that the individual processes are ⁠ M {\displaystyle M} ⁠th- respectively ⁠ N {\displaystyle N} ⁠th-order stationary.[1]: 159

## Other terminology

The terminology used for types of stationarity other than strict stationarity can be rather mixed. Some examples follow.

- [Priestley](/source/Maurice_Priestley) uses **stationary up to order** ⁠ m {\displaystyle m} ⁠ if conditions similar to those given here for wide sense stationarity apply relating to moments up to order ⁠ m {\displaystyle m} ⁠.[3][4] Thus wide sense stationarity would be equivalent to "stationary to order 2", which is different from the definition of second-order stationarity given here.

- [Honarkhah](https://en.wikipedia.org/w/index.php?title=Mehrdad_Honarkhah&action=edit&redlink=1) and [Caers](/source/Jef_Caers) also use the assumption of stationarity in the context of multiple-point geostatistics, where higher ⁠ n {\displaystyle n} ⁠-point statistics are assumed to be stationary in the spatial domain.[5]

## Techniques to stationarize a non-stationary process

In time series analysis and stochastic processes, stationarizing a time series is a crucial preprocessing step aimed at transforming a non-stationary process into a stationary one. Several techniques exist for achieving this, depending on the type and order of non-stationarity present. For first-order non-stationarity, where the mean of the process varies over time, differencing is a common and effective method: it transforms the series by subtracting each value from its predecessor, thus stabilizing the mean. For non-stationarities up to the second order, time-frequency analysis (e.g., [Wavelet transform](/source/Wavelet_transform), [Wigner distribution function](/source/Wigner_distribution_function), or [Short-time Fourier transform](/source/Short-time_Fourier_transform)) can be employed to isolate and suppress time-localized, nonstationary spectral components. Additionally, surrogate data methods can be used to construct strictly stationary versions of the original time series. One of the ways for identifying non-stationary times series is the [ACF](/source/Autocorrelation) plot. Sometimes, patterns will be more visible in the ACF plot than in the original time series; however, this is not always the case.[6]

The choice of method for time series stationarization depends on the nature of the non-stationarity and the goals of the analysis, especially when building models that require strict stationarity assumptions, such as ARMA or spectral-based techniques. More details on some time series stationarization methods are presented below.

### Stationarization by means of differencing

One way to make some time series first-order stationary is to compute the differences between consecutive observations. This is known as [differencing](/source/Unit_root). Differencing can help stabilize the mean of a time series by removing changes in the level of a time series, and so eliminating trends. This can also remove seasonality, if differences are taken appropriately (e.g. differencing observations 1 year apart to remove a yearly trend). Transformations such as logarithms can help to stabilize the variance of a time series.

### Stationarization by means of the surrogate method

The surrogate method for stationarization[7] works by generating a new time series that preserves certain statistical properties of the original series while removing its nonstationary components.[8][9][10] A common approach is to apply the Fourier Transform to the original time series to obtain its magnitude and phase spectra. The magnitude spectrum, which determines the power distribution across frequencies, is retained to preserve the global autocorrelation structure. The phase spectrum, which encodes the temporal alignment of frequency components and is often responsible for time-dependent dynamics in the time series (like non-stationarities), is then randomized, typically by replacing it with a set of random phases drawn uniformly from [ − π , π ] {\displaystyle [-\pi ,\pi ]} while enforcing conjugate symmetry to ensure a real-valued inverse. Applying the inverse Fourier Transform to the modified spectra yields a strictly stationary surrogate time series:[11] one with the same power spectrum as the original but lacking the temporal structures that caused non-stationarity. This technique is often used in hypothesis tests for probing the stationarity property.[8][10][12][13]

## See also

- [Lévy process](/source/L%C3%A9vy_process)

- [Stationary ergodic process](/source/Stationary_ergodic_process)

- [Wiener–Khinchin theorem](/source/Wiener%E2%80%93Khinchin_theorem)

- [Ergodicity](/source/Ergodicity)

- [Statistical regularity](/source/Statistical_regularity)

- [Autocorrelation](/source/Autocorrelation)

- [Whittle likelihood](/source/Whittle_likelihood)

## References

1. ^ [***a***](#cite_ref-KunIlPark_1-0) [***b***](#cite_ref-KunIlPark_1-1) [***c***](#cite_ref-KunIlPark_1-2) [***d***](#cite_ref-KunIlPark_1-3) [***e***](#cite_ref-KunIlPark_1-4) [***f***](#cite_ref-KunIlPark_1-5) [***g***](#cite_ref-KunIlPark_1-6) Park, Kun Il (2018). *Fundamentals of Probability and Stochastic Processes with Applications to Communications*. Springer. [ISBN](/source/ISBN_(identifier)) [978-3-319-68074-3](https://en.wikipedia.org/wiki/Special:BookSources/978-3-319-68074-3).

1. ^ [***a***](#cite_ref-Florescu2014_2-0) [***b***](#cite_ref-Florescu2014_2-1) Ionut Florescu (7 November 2014). *Probability and Stochastic Processes*. John Wiley & Sons. [ISBN](/source/ISBN_(identifier)) [978-1-118-59320-2](https://en.wikipedia.org/wiki/Special:BookSources/978-1-118-59320-2).

1. **[^](#cite_ref-3)** Priestley, M. B. (1981). *Spectral Analysis and Time Series*. Academic Press. [ISBN](/source/ISBN_(identifier)) [0-12-564922-3](https://en.wikipedia.org/wiki/Special:BookSources/0-12-564922-3).

1. **[^](#cite_ref-4)** Priestley, M. B. (1988). [*Non-linear and Non-stationary Time Series Analysis*](https://archive.org/details/nonlinearnonstat0000prie). Academic Press. [ISBN](/source/ISBN_(identifier)) [0-12-564911-8](https://en.wikipedia.org/wiki/Special:BookSources/0-12-564911-8).

1. **[^](#cite_ref-5)** Honarkhah, M.; Caers, J. (2010). "Stochastic Simulation of Patterns Using Distance-Based Pattern Modeling". *Mathematical Geosciences*. **42** (5): 487–517. [Bibcode](/source/Bibcode_(identifier)):[2010MatGe..42..487H](https://ui.adsabs.harvard.edu/abs/2010MatGe..42..487H). [doi](/source/Doi_(identifier)):[10.1007/s11004-010-9276-7](https://doi.org/10.1007%2Fs11004-010-9276-7).

1. **[^](#cite_ref-6)** Hyndman, Rob J.; Athanasopoulos, George. "8.1 Stationarity and differencing". [*Forecasting: Principles and Practice*](https://www.otexts.org/fpp/8/1) (2nd ed.). OTexts. Retrieved 2016-05-18.

1. **[^](#cite_ref-7)** Pierre Borgnat and Patrick Flandrin. (2009). Stationarization via surrogates. Journal of Statistical Mechanics: Theory and Experiment, vol. 2009, n. 1, [https://iopscience.iop.org/article/10.1088/1742-5468/2009/01/P01001](https://iopscience.iop.org/article/10.1088/1742-5468/2009/01/P01001)

1. ^ [***a***](#cite_ref-ieeexplore.ieee.org_8-0) [***b***](#cite_ref-ieeexplore.ieee.org_8-1) Pierre Borgnat et al. (2010). Testing Stationarity With Surrogates: A Time-Frequency Approach. IEEE Transactions on Signal Processing, vol. 58, n. 7, pp. 3459-3470 [https://ieeexplore.ieee.org/document/5419113](https://ieeexplore.ieee.org/document/5419113)

1. **[^](#cite_ref-9)** Pierre Borgnat et al. (2011). Transitional Surrogates. 2011 IEEE International Conference on Acoustics, Speech and Signal Processing, pp. 3600-3603 [https://ieeexplore.ieee.org/document/5946257](https://ieeexplore.ieee.org/document/5946257)

1. ^ [***a***](#cite_ref-Souza_2019_10-0) [***b***](#cite_ref-Souza_2019_10-1) Douglas Baptista de Souza et al. (2019). An Improved Stationarity Test Based on Surrogates. IEEE Signal Processing Letters, vol. 26, n. 10, pp. 1431-1435 [https://ieeexplore.ieee.org/abstract/document/8777090](https://ieeexplore.ieee.org/abstract/document/8777090)

1. **[^](#cite_ref-11)** Cédric Richard et al. (2010). Statistical hypothesis testing with time-frequency surrogates to check signal stationarity. 2010 IEEE International Conference on Acoustics, Speech and Signal Processing, pp. 3666-3669 [https://ieeexplore.ieee.org/document/5495887](https://ieeexplore.ieee.org/document/5495887)

1. **[^](#cite_ref-12)** Douglas Baptista de Souza et al. (2012). A modified time-frequency method for testing wide-sense stationarity. 2012 IEEE International Conference on Acoustics, Speech and Signal Processing, pp. 3409-3412 [https://ieeexplore.ieee.org/abstract/document/6288648](https://ieeexplore.ieee.org/abstract/document/6288648)

1. **[^](#cite_ref-13)** Jun Xiao et al. (2007). Testing Stationarity with Surrogates - A One-Class SVM Approach. 2007 IEEE/SP 14th Workshop on Statistical Signal Processing, pp. 720-724 [https://ieeexplore.ieee.org/document/4301353](https://ieeexplore.ieee.org/document/4301353)

## Further reading

- Enders, Walter (2010). *Applied Econometric Time Series* (Third ed.). New York: Wiley. pp. 53–57. [ISBN](/source/ISBN_(identifier)) [978-0-470-50539-7](https://en.wikipedia.org/wiki/Special:BookSources/978-0-470-50539-7).

- Jestrovic, I.; Coyle, J. L.; Sejdic, E (2015). ["The effects of increased fluid viscosity on stationary characteristics of EEG signal in healthy adults"](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4253861). *Brain Research*. **1589**: 45–53. [doi](/source/Doi_(identifier)):[10.1016/j.brainres.2014.09.035](https://doi.org/10.1016%2Fj.brainres.2014.09.035). [PMC](/source/PMC_(identifier)) [4253861](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4253861). [PMID](/source/PMID_(identifier)) [25245522](https://pubmed.ncbi.nlm.nih.gov/25245522).

- Hyndman, Athanasopoulos (2013). Forecasting: Principles and Practice. Otexts. [https://www.otexts.org/fpp/8/1](https://www.otexts.org/fpp/8/1)

## External links

- [Spectral decomposition of a random function (Springer)](https://encyclopediaofmath.org/wiki/Spectral_decomposition_of_a_random_function)

v t e Stochastic processes Discrete time Bernoulli process Branching process Chinese restaurant process Galton–Watson process Independent and identically distributed random variables Markov chain Moran process Random walk Loop-erased Self-avoiding Biased Maximal entropy Continuous time Additive process Airy process Bessel process Birth–death process pure birth Brownian motion Bridge Dyson Excursion Fractional Geometric Meander Cauchy process Contact process Continuous-time random walk Cox process Diffusion process Empirical process Feller process Fleming–Viot process Gamma process Geometric process Hawkes process Hunt process Interacting particle systems Itô diffusion Itô process Jump diffusion Jump process Lévy process Local time Markov additive process McKean–Vlasov process Ornstein–Uhlenbeck process Poisson process Compound Non-homogeneous Quasimartingale Schramm–Loewner evolution Semimartingale Sigma-martingale Stable process Superprocess Telegraph process Variance gamma process Wiener process Wiener sausage Both Branching process Gaussian process Hidden Markov model (HMM) Markov process Martingale Differences Local Sub- Super- Random dynamical system Regenerative process Renewal process Stochastic chains with memory of variable length White noise Fields and other Dirichlet process Gaussian random field Gibbs measure Hopfield model Ising model Potts model Boolean network Markov random field Percolation Pitman–Yor process Point process Cox Determinantal Poisson Random field Random graph Time series models Autoregressive conditional heteroskedasticity (ARCH) model Autoregressive integrated moving average (ARIMA) model Autoregressive (AR) model Autoregressive moving-average (ARMA) model Generalized autoregressive conditional heteroskedasticity (GARCH) model Moving-average (MA) model Financial models Binomial options pricing model Black–Derman–Toy Black–Karasinski Black–Scholes Chan–Karolyi–Longstaff–Sanders (CKLS) Chen Constant elasticity of variance (CEV) Cox–Ingersoll–Ross (CIR) Garman–Kohlhagen Heath–Jarrow–Morton (HJM) Heston Ho–Lee Hull–White Korn-Kreer-Lenssen LIBOR market Rendleman–Bartter SABR volatility Vašíček Wilkie Actuarial models Bühlmann Cramér–Lundberg Risk process Sparre–Anderson Queueing models Bulk Fluid Generalized queueing network M/G/1 M/M/1 M/M/c Properties Càdlàg paths Continuous Continuous paths Ergodic Exchangeable Feller-continuous Gauss–Markov Markov Mixing Piecewise-deterministic Predictable Progressively measurable Self-similar Stationary Time-reversible Limit theorems Central limit theorem Donsker's theorem Doob's martingale convergence theorems Ergodic theorem Fisher–Tippett–Gnedenko theorem Large deviation principle Law of large numbers (weak/strong) Law of the iterated logarithm Maximal ergodic theorem Sanov's theorem Zero–one laws (Blumenthal, Borel–Cantelli, Engelbert–Schmidt, Hewitt–Savage, Kolmogorov, Lévy) Inequalities Burkholder–Davis–Gundy Doob's martingale Doob's upcrossing Kunita–Watanabe Marcinkiewicz–Zygmund Tools Cameron–Martin theorem Convergence of random variables Doléans-Dade exponential Doob decomposition theorem Doob–Meyer decomposition theorem Doob's optional stopping theorem Dynkin's formula Feynman–Kac formula Filtration Girsanov theorem Infinitesimal generator Itô integral Itô's lemma Kolmogorov continuity theorem Kolmogorov extension theorem Kosambi–Karhunen–Loève theorem Lévy–Prokhorov metric Malliavin calculus Martingale representation theorem Optional stopping theorem Prokhorov's theorem Quadratic variation Reflection principle Skorokhod integral Skorokhod's representation theorem Skorokhod space Snell envelope Stochastic differential equation Tanaka Stopping time Stratonovich integral Uniform integrability Usual hypotheses Wiener space Classical Abstract Disciplines Actuarial mathematics Control theory Econometrics Ergodic theory Extreme value theory (EVT) Large deviations theory Mathematical finance Mathematical statistics Probability theory Queueing theory Renewal theory Ruin theory Signal processing Statistics Stochastic analysis Time series analysis Machine learning List of topics Category

v t e Statistics Outline Index Descriptive statistics Continuous data Center Mean Arithmetic Arithmetic-Geometric Contraharmonic Cubic Generalized/power Geometric Harmonic Heronian Heinz Lehmer Median Mode Dispersion Average absolute deviation Coefficient of variation Interquartile range Percentile Range Standard deviation Variance Shape Central limit theorem Moments Kurtosis L-moments Skewness Count data Index of dispersion Summary tables Contingency table Frequency distribution Grouped data Dependence Partial correlation Pearson product-moment correlation Rank correlation Kendall's τ Spearman's ρ Scatter plot Graphics Bar chart Biplot Box plot Control chart Correlogram Fan chart Forest plot Histogram Pie chart Q–Q plot Radar chart Run chart Scatter plot Stem-and-leaf display Violin plot Heatmap Scatter Plot Matrix ECDF plot Line chart Statistical data processing Transformations Data transformation Log transformation Power transform Box–Cox transformation Yeo–Johnson transformation Variance-stabilizing transformation Anscombe transform Fisher transformation Scaling and normalization Feature scaling Normalization Standardization (z-score) Min–max normalization Unit vector normalization Data cleaning Data cleaning Outlier Winsorizing Truncation Missing data Data reduction Dimensionality reduction Principal component analysis Factor analysis Time-series preprocessing Differencing Detrending Seasonal adjustment Stationarity transformation Data collection Study design Effect size Missing data Optimal design Population Replication Sample size determination Statistic Statistical power Survey methodology Sampling Cluster Stratified Opinion poll Questionnaire Standard error Controlled experiments Blocking Factorial experiment Interaction Random assignment Randomized controlled trial Randomized experiment Scientific control Adaptive designs Adaptive clinical trial Stochastic approximation Up-and-down designs Observational studies Cohort study Cross-sectional study Natural experiment Quasi-experiment Statistical inference Statistical theory Population Statistic Probability distribution Sampling distribution Order statistic Empirical distribution Density estimation Statistical model Model specification Lp space Parameter location scale shape Parametric family Likelihood (monotone) Location–scale family Exponential family Completeness Sufficiency Statistical functional Bootstrap U V Optimal decision loss function Efficiency Statistical distance divergence Asymptotics Robustness Frequentist inference Point estimation Estimating equations Maximum likelihood Method of moments M-estimator Minimum distance Unbiased estimators Mean-unbiased minimum-variance Rao–Blackwellization Lehmann–Scheffé theorem Median unbiased Plug-in Interval estimation Confidence interval Pivot Likelihood interval Prediction interval Tolerance interval Resampling Bootstrap Jackknife Testing hypotheses 1- & 2-tails Power Uniformly most powerful test Permutation test Randomization test Multiple comparisons Parametric tests Likelihood-ratio Score/Lagrange multiplier Wald Specific tests Z-test (normal) Student's t-test F-test Goodness of fit Chi-squared G-test Kolmogorov–Smirnov Anderson–Darling Lilliefors Jarque–Bera Normality (Shapiro–Wilk) Likelihood-ratio test Model selection Cross validation AIC BIC Rank statistics Sign Sample median Signed rank (Wilcoxon) Hodges–Lehmann estimator Rank sum (Mann–Whitney) Nonparametric anova 1-way (Kruskal–Wallis) 2-way (Friedman) Ordered alternative (Jonckheere–Terpstra) Van der Waerden test Bayesian inference Bayesian probability prior posterior Credible interval Bayes factor Bayesian estimator Maximum posterior estimator Correlation Regression analysis Correlation Pearson product-moment Partial correlation Confounding variable Coefficient of determination Regression analysis Errors and residuals Regression validation Mixed effects models Simultaneous equations models Multivariate adaptive regression splines (MARS) Template:Least squares and regression analysis Linear regression Simple linear regression Ordinary least squares General linear model Bayesian regression Non-standard predictors Nonlinear regression Nonparametric Semiparametric Isotonic Robust Homoscedasticity and Heteroscedasticity Generalized linear model Exponential families Logistic (Bernoulli) / Binomial / Poisson regressions Partition of variance Analysis of variance (ANOVA, anova) Analysis of covariance Multivariate ANOVA Degrees of freedom Categorical / multivariate / time-series / survival analysis Categorical Cohen's kappa Contingency table Graphical model Log-linear model McNemar's test Cochran–Mantel–Haenszel statistics Multivariate Regression Manova Principal components Canonical correlation Discriminant analysis Cluster analysis Classification Structural equation model Factor analysis Multivariate distributions Elliptical distributions Normal Time-series General Decomposition Trend Stationarity Seasonal adjustment Exponential smoothing Cointegration Structural break Granger causality Specific tests Dickey–Fuller Johansen Q-statistic (Ljung–Box) Durbin–Watson Breusch–Godfrey Time domain Autocorrelation (ACF) partial (PACF) Cross-correlation (XCF) ARMA model ARIMA model (Box–Jenkins) Autoregressive conditional heteroskedasticity (ARCH) Vector autoregression (VAR) (Autoregressive model (AR)) Frequency domain Spectral density estimation Fourier analysis Least-squares spectral analysis Wavelet Whittle likelihood Survival Survival function Kaplan–Meier estimator (product limit) Proportional hazards models Accelerated failure time (AFT) model First hitting time Hazard function Nelson–Aalen estimator Test Log-rank test Applications Biostatistics Bioinformatics Clinical trials / studies Epidemiology Medical statistics Engineering statistics Chemometrics Methods engineering Probabilistic design Process / quality control Reliability System identification Social statistics Actuarial science Census Crime statistics Demography Econometrics Jurimetrics National accounts Official statistics Population statistics Psychometrics Spatial statistics Cartography Environmental statistics Geographic information system Geostatistics Kriging Category Mathematics portal Commons WikiProject

---
Adapted from the Wikipedia article [Stationary process](https://en.wikipedia.org/wiki/Stationary_process) by Wikipedia contributors ([contributor history](https://en.wikipedia.org/wiki/Stationary_process?action=history)). Available under [Creative Commons Attribution-ShareAlike 4.0 International](https://creativecommons.org/licenses/by-sa/4.0/). Changes may have been made.