In statistics, propagation of uncertainty (or propagation of error) is the effect of variables' uncertainties (or errors, more specifically random errors) on the uncertainty of a function based on them. When the variables are the values of experimental measurements they have uncertainties due to measurement limitations (e.g., instrument precision) which propagate due to the combination of variables in the function.

The uncertainty u can be expressed in a number of ways. It may be defined by the absolute error Δx. Uncertainties can also be defined by the relative error (Δx)/x, which is usually written as a percentage. Most commonly, the uncertainty on a quantity is quantified in terms of the standard deviation, σ, which is the positive square root of the variance. The value of a quantity and its error are then expressed as an interval x ± u. However, the most general way of characterizing uncertainty is by specifying its probability distribution. If the probability distribution of the variable is known or can be assumed, in theory it is possible to get any of its statistics. In particular, it is possible to derive confidence limits to describe the region within which the true value of the variable may be found. For example, the 68% confidence limits for a one-dimensional variable belonging to a normal distribution are approximately ± one standard deviation σ from the central value x, which means that the region x ± σ will cover the true value in roughly 68% of cases.

If the uncertainties are correlated then covariance must be taken into account. Correlation can arise from two different sources. First, the measurement errors may be correlated. Second, when the underlying values are correlated across a population, the uncertainties in the group averages will be correlated.^[1]

In a general context where a nonlinear function modifies the uncertain parameters (correlated or not), the standard tools to propagate uncertainty, and infer resulting quantity probability distribution/statistics, are sampling techniques from the Monte Carlo method family.^[2] For very expansive data or complex functions, the calculation of the error propagation may be very expansive so that a surrogate model^[3] or a parallel computing strategy^[4][5][6] may be necessary.

In some particular cases, the uncertainty propagation calculation can be done through simplistic algebraic procedures. Some of these scenarios are described below.

Let ${\displaystyle \{f_{k}(x_{1},x_{2},\dots ,x_{n})\}}$ be a set of m functions, which are linear combinations of ${\displaystyle n}$ variables ${\displaystyle x_{1},x_{2},\dots ,x_{n}}$ with combination coefficients ${\displaystyle A_{k1},A_{k2},\dots ,A_{kn},(k=1,\dots ,m)}$:

$${\displaystyle f_{k}=\sum _{i=1}^{n}A_{ki}x_{i},}$$

or in matrix notation,

$${\displaystyle \mathbf {f} =\mathbf {A} \mathbf {x} .}$$

Also let the variance–covariance matrix of x = (x₁, ..., x_n) be denoted by ${\displaystyle {\boldsymbol {\Sigma }}^{x}}$ and let the mean value be denoted by ${\displaystyle {\boldsymbol {\mu }}}$:

$${\displaystyle {\boldsymbol {\Sigma }}^{x}=E[(\mathbf {x} -{\boldsymbol {\mu }})\otimes (\mathbf {x} -{\boldsymbol {\mu }})]={\begin{pmatrix}\sigma _{1}^{2}&\sigma _{12}&\sigma _{13}&\cdots \\\sigma _{21}&\sigma _{2}^{2}&\sigma _{23}&\cdots \\\sigma _{31}&\sigma _{32}&\sigma _{3}^{2}&\cdots \\\vdots &\vdots &\vdots &\ddots \end{pmatrix}}={\begin{pmatrix}{\Sigma }_{11}^{x}&{\Sigma }_{12}^{x}&{\Sigma }_{13}^{x}&\cdots \\{\Sigma }_{21}^{x}&{\Sigma }_{22}^{x}&{\Sigma }_{23}^{x}&\cdots \\{\Sigma }_{31}^{x}&{\Sigma }_{32}^{x}&{\Sigma }_{33}^{x}&\cdots \\\vdots &\vdots &\vdots &\ddots \end{pmatrix}}.}$$

${\displaystyle \otimes }$ is the outer product.

Then, the variance–covariance matrix ${\displaystyle {\boldsymbol {\Sigma }}^{f}}$ of f is given by

$${\displaystyle {\boldsymbol {\Sigma }}^{f}=E[(\mathbf {f} -E[\mathbf {f} ])\otimes (\mathbf {f} -E[\mathbf {f} ])]=E[\mathbf {A} (\mathbf {x} -{\boldsymbol {\mu }})\otimes \mathbf {A} (\mathbf {x} -{\boldsymbol {\mu }})]=\mathbf {A} E[(\mathbf {x} -{\boldsymbol {\mu }})\otimes (\mathbf {x} -{\boldsymbol {\mu }})]\mathbf {A} ^{\mathrm {T} }=\mathbf {A} {\boldsymbol {\Sigma }}^{x}\mathbf {A} ^{\mathrm {T} }.}$$

In component notation, the equation

$${\displaystyle {\boldsymbol {\Sigma }}^{f}=\mathbf {A} {\boldsymbol {\Sigma }}^{x}\mathbf {A} ^{\mathrm {T} }}$$

reads

$${\displaystyle \Sigma _{ij}^{f}=\sum _{k}^{n}\sum _{l}^{n}A_{ik}{\Sigma }_{kl}^{x}A_{jl}.}$$

This is the most general expression for the propagation of error from one set of variables onto another. When the errors on x are uncorrelated, the general expression simplifies to

$${\displaystyle \Sigma _{ij}^{f}=\sum _{k}^{n}A_{ik}\Sigma _{k}^{x}A_{jk},}$$

where ${\displaystyle \Sigma _{k}^{x}=\sigma _{x_{k}}^{2}}$ is the variance of k-th element of the x vector. Note that even though the errors on x may be uncorrelated, the errors on f are in general correlated; in other words, even if ${\displaystyle {\boldsymbol {\Sigma }}^{x}}$ is a diagonal matrix, ${\displaystyle {\boldsymbol {\Sigma }}^{f}}$ is in general a full matrix.

The general expressions for a scalar-valued function f are a little simpler (here a is a row vector):

$${\displaystyle f=\sum _{i}^{n}a_{i}x_{i}=\mathbf {ax} ,}$$

$${\displaystyle \sigma _{f}^{2}=\sum _{i}^{n}\sum _{j}^{n}a_{i}\Sigma _{ij}^{x}a_{j}=\mathbf {a} {\boldsymbol {\Sigma }}^{x}\mathbf {a} ^{\mathrm {T} }.}$$

Each covariance term ${\displaystyle \sigma _{ij}}$ can be expressed in terms of the correlation coefficient ${\displaystyle \rho _{ij}}$ by ${\displaystyle \sigma _{ij}=\rho _{ij}\sigma _{i}\sigma _{j}}$, so that an alternative expression for the variance of f is

$${\displaystyle \sigma _{f}^{2}=\sum _{i}^{n}a_{i}^{2}\sigma _{i}^{2}+\sum _{i}^{n}\sum _{j(j\neq i)}^{n}a_{i}a_{j}\rho _{ij}\sigma _{i}\sigma _{j}.}$$

In the case that the variables in x are uncorrelated, this simplifies further to

$${\displaystyle \sigma _{f}^{2}=\sum _{i}^{n}a_{i}^{2}\sigma _{i}^{2}.}$$

In the simple case of identical coefficients and variances, we find

$${\displaystyle \sigma _{f}={\sqrt {n}}\,|a|\sigma .}$$

For the arithmetic mean, ${\displaystyle a=1/n}$, the result is the standard error of the mean:

$${\displaystyle \sigma _{f}={\frac {\sigma }{\sqrt {n}}}.}$$

When f is a set of non-linear combination of the variables x, an interval propagation could be performed in order to compute intervals which contain all consistent values for the variables. In a probabilistic approach, the function f must usually be linearised by approximation to a first-order Taylor series expansion, though in some cases, exact formulae can be derived that do not depend on the expansion as is the case for the exact variance of products.^[7] The Taylor expansion would be:

$${\displaystyle f_{k}\approx f_{k}^{0}+\sum _{i}^{n}{\frac {\partial f_{k}}{\partial {x_{i}}}}x_{i}}$$

where ${\displaystyle \partial f_{k}/\partial x_{i}}$ denotes the partial derivative of f_k with respect to the i-th variable, evaluated at the mean value of all components of vector x. Or in matrix notation,

$${\displaystyle \mathrm {f} \approx \mathrm {f} ^{0}+\mathrm {J} \mathrm {x} \,}$$

where J is the Jacobian matrix. Since f⁰ is a constant it does not contribute to the error on f. Therefore, the propagation of error follows the linear case, above, but replacing the linear coefficients, A_ki and A_kj by the partial derivatives, ${\displaystyle {\frac {\partial f_{k}}{\partial x_{i}}}}$ and ${\displaystyle {\frac {\partial f_{k}}{\partial x_{j}}}}$. In matrix notation,^[8]

$${\displaystyle \mathrm {\Sigma } ^{\mathrm {f} }=\mathrm {J} \mathrm {\Sigma } ^{\mathrm {x} }\mathrm {J} ^{\top }.}$$

That is, the Jacobian of the function is used to transform the rows and columns of the variance-covariance matrix of the argument. Note this is equivalent to the matrix expression for the linear case with ${\displaystyle \mathrm {J=A} }$.

Simplification

Neglecting correlations or assuming independent variables yields a common formula among engineers and experimental scientists to calculate error propagation, the variance formula:^[9]

$${\displaystyle s_{f}={\sqrt {\left({\frac {\partial f}{\partial x}}\right)^{2}s_{x}^{2}+\left({\frac {\partial f}{\partial y}}\right)^{2}s_{y}^{2}+\left({\frac {\partial f}{\partial z}}\right)^{2}s_{z}^{2}+\cdots }}}$$

where ${\displaystyle s_{f}}$ represents the standard deviation of the function ${\displaystyle f}$, ${\displaystyle s_{x}}$ represents the standard deviation of ${\displaystyle x}$, ${\displaystyle s_{y}}$ represents the standard deviation of ${\displaystyle y}$, and so forth.

It is important to note that this formula is based on the linear characteristics of the gradient of ${\displaystyle f}$ and therefore it is a good estimation for the standard deviation of ${\displaystyle f}$ as long as ${\displaystyle s_{x},s_{y},s_{z},\ldots }$ are small enough. Specifically, the linear approximation of ${\displaystyle f}$ has to be close to ${\displaystyle f}$ inside a neighbourhood of radius ${\displaystyle s_{x},s_{y},s_{z},\ldots }$.^[10]

Example

Any non-linear differentiable function, ${\displaystyle f(a,b)}$, of two variables, ${\displaystyle a}$ and ${\displaystyle b}$, can be expanded as

$${\displaystyle f\approx f^{0}+{\frac {\partial f}{\partial a}}a+{\frac {\partial f}{\partial b}}b}$$

now, taking variance on both sides, and using the formula^[11] for variance of a linear combination of variables:

$${\displaystyle \operatorname {Var} (aX+bY)=a^{2}\operatorname {Var} (X)+b^{2}\operatorname {Var} (Y)+2ab\operatorname {Cov} (X,Y)}$$

hence:

$${\displaystyle \sigma _{f}^{2}\approx \left|{\frac {\partial f}{\partial a}}\right|^{2}\sigma _{a}^{2}+\left|{\frac {\partial f}{\partial b}}\right|^{2}\sigma _{b}^{2}+2{\frac {\partial f}{\partial a}}{\frac {\partial f}{\partial b}}\sigma _{ab}}$$

where ${\displaystyle \sigma _{f}}$ is the standard deviation of the function ${\displaystyle f}$, ${\displaystyle \sigma _{a}}$ is the standard deviation of ${\displaystyle a}$, ${\displaystyle \sigma _{b}}$ is the standard deviation of ${\displaystyle b}$ and ${\displaystyle \sigma _{ab}=\sigma _{a}\sigma _{b}\rho _{ab}}$ is the covariance between ${\displaystyle a}$ and ${\displaystyle b}$.

In the particular case that ${\displaystyle f=ab}$, ${\displaystyle {\frac {\partial f}{\partial a}}=b}$, ${\displaystyle {\frac {\partial f}{\partial b}}=a}$. Then

$${\displaystyle \sigma _{f}^{2}\approx b^{2}\sigma _{a}^{2}+a^{2}\sigma _{b}^{2}+2ab\,\sigma _{ab}}$$

or

$${\displaystyle \left({\frac {\sigma _{f}}{f}}\right)^{2}\approx \left({\frac {\sigma _{a}}{a}}\right)^{2}+\left({\frac {\sigma _{b}}{b}}\right)^{2}+2\left({\frac {\sigma _{a}}{a}}\right)\left({\frac {\sigma _{b}}{b}}\right)\rho _{ab}}$$

where ${\displaystyle \rho _{ab}}$ is the correlation between ${\displaystyle a}$ and ${\displaystyle b}$.

When the variables ${\displaystyle a}$ and ${\displaystyle b}$ are uncorrelated, ${\displaystyle \rho _{ab}=0}$. Then

$${\displaystyle \left({\frac {\sigma _{f}}{f}}\right)^{2}\approx \left({\frac {\sigma _{a}}{a}}\right)^{2}+\left({\frac {\sigma _{b}}{b}}\right)^{2}.}$$

This table shows the variances and standard deviations of simple functions of the real variables ${\displaystyle A,B\!}$, with standard deviations ${\displaystyle \sigma _{A},\sigma _{B},\,}$ covariance ${\displaystyle \sigma _{AB}=\rho _{AB}\sigma _{A}\sigma _{B}\,}$, and correlation ${\displaystyle \rho _{AB}}$. The real-valued coefficients ${\displaystyle a}$ and ${\displaystyle b}$ are assumed exactly known (deterministic), i.e., ${\displaystyle \sigma _{a}=\sigma _{b}=0}$.

In the right-hand columns of the table, ${\displaystyle A}$ and ${\displaystyle B}$ are expectation values and ${\displaystyle f}$ is the value of the function calculated at those values.

More information , ...

Function	Variance	Standard Deviation
${\displaystyle f=aA\,}$	${\displaystyle \sigma _{f}^{2}=a^{2}\sigma _{A}^{2}}$	${\displaystyle \sigma _{f}=|a|\sigma _{A}}$
${\displaystyle f=A+B\,}$	${\displaystyle \sigma _{f}^{2}=\sigma _{A}^{2}+\sigma _{B}^{2}+2\sigma _{AB}}$	${\displaystyle \sigma _{f}={\sqrt {\sigma _{A}^{2}+\sigma _{B}^{2}+2\sigma _{AB}}}}$
${\displaystyle f=A-B\,}$	${\displaystyle \sigma _{f}^{2}=\sigma _{A}^{2}+\sigma _{B}^{2}-2\sigma _{AB}}$	${\displaystyle \sigma _{f}={\sqrt {\sigma _{A}^{2}+\sigma _{B}^{2}-2\sigma _{AB}}}}$
${\displaystyle f=aA+bB\,}$	${\displaystyle \sigma _{f}^{2}=a^{2}\sigma _{A}^{2}+b^{2}\sigma _{B}^{2}+2ab\,\sigma _{AB}}$	${\displaystyle \sigma _{f}={\sqrt {a^{2}\sigma _{A}^{2}+b^{2}\sigma _{B}^{2}+2ab\,\sigma _{AB}}}}$
${\displaystyle f=aA-bB\,}$	${\displaystyle \sigma _{f}^{2}=a^{2}\sigma _{A}^{2}+b^{2}\sigma _{B}^{2}-2ab\,\sigma _{AB}}$	${\displaystyle \sigma _{f}={\sqrt {a^{2}\sigma _{A}^{2}+b^{2}\sigma _{B}^{2}-2ab\,\sigma _{AB}}}}$
${\displaystyle f=AB\,}$	${\displaystyle \sigma _{f}^{2}\approx f^{2}\left[\left({\frac {\sigma _{A}}{A}}\right)^{2}+\left({\frac {\sigma _{B}}{B}}\right)^{2}+2{\frac {\sigma _{AB}}{AB}}\right]}$[15][16]	${\displaystyle \sigma _{f}\approx \left|f\right|{\sqrt {\left({\frac {\sigma _{A}}{A}}\right)^{2}+\left({\frac {\sigma _{B}}{B}}\right)^{2}+2{\frac {\sigma _{AB}}{AB}}}}}$
${\displaystyle f={\frac {A}{B}}\,}$	${\displaystyle \sigma _{f}^{2}\approx f^{2}\left[\left({\frac {\sigma _{A}}{A}}\right)^{2}+\left({\frac {\sigma _{B}}{B}}\right)^{2}-2{\frac {\sigma _{AB}}{AB}}\right]}$[17]	${\displaystyle \sigma _{f}\approx \left|f\right|{\sqrt {\left({\frac {\sigma _{A}}{A}}\right)^{2}+\left({\frac {\sigma _{B}}{B}}\right)^{2}-2{\frac {\sigma _{AB}}{AB}}}}}$
${\displaystyle f={\frac {A}{A+B}}\,}$	${\displaystyle \sigma _{f}^{2}\approx {\frac {f^{2}}{\left(A+B\right)^{2}}}\left({\frac {B^{2}}{A^{2}}}\sigma _{A}^{2}+\sigma _{B}^{2}-2{\frac {B}{A}}\sigma _{AB}\right)}$	${\displaystyle \sigma _{f}\approx \left|{\frac {f}{A+B}}\right|{\sqrt {{\frac {B^{2}}{A^{2}}}\sigma _{A}^{2}+\sigma _{B}^{2}-2{\frac {B}{A}}\sigma _{AB}}}}$
${\displaystyle f=aA^{b}\,}$	${\displaystyle \sigma _{f}^{2}\approx \left({a}{b}{A}^{b-1}{\sigma _{A}}\right)^{2}=\left({\frac {{f}{b}{\sigma _{A}}}{A}}\right)^{2}}$	${\displaystyle \sigma _{f}\approx \left|{a}{b}{A}^{b-1}{\sigma _{A}}\right|=\left|{\frac {{f}{b}{\sigma _{A}}}{A}}\right|}$
${\displaystyle f=a\ln(bA)\,}$	${\displaystyle \sigma _{f}^{2}\approx \left(a{\frac {\sigma _{A}}{A}}\right)^{2}}$[18]	${\displaystyle \sigma _{f}\approx \left|a{\frac {\sigma _{A}}{A}}\right|}$
${\displaystyle f=a\log _{10}(bA)\,}$	${\displaystyle \sigma _{f}^{2}\approx \left(a{\frac {\sigma _{A}}{A\ln(10)}}\right)^{2}}$[18]	${\displaystyle \sigma _{f}\approx \left|a{\frac {\sigma _{A}}{A\ln(10)}}\right|}$
${\displaystyle f=ae^{bA}\,}$	${\displaystyle \sigma _{f}^{2}\approx f^{2}\left(b\sigma _{A}\right)^{2}}$[19]	${\displaystyle \sigma _{f}\approx \left|f\right|\left|\left(b\sigma _{A}\right)\right|}$
${\displaystyle f=a^{bA}\,}$	${\displaystyle \sigma _{f}^{2}\approx f^{2}(b\ln(a)\sigma _{A})^{2}}$	${\displaystyle \sigma _{f}\approx \left|f\right|\left|b\ln(a)\sigma _{A}\right|}$
${\displaystyle f=a\sin(bA)\,}$	${\displaystyle \sigma _{f}^{2}\approx \left[ab\cos(bA)\sigma _{A}\right]^{2}}$	${\displaystyle \sigma _{f}\approx \left|ab\cos(bA)\sigma _{A}\right|}$
${\displaystyle f=a\cos \left(bA\right)\,}$	${\displaystyle \sigma _{f}^{2}\approx \left[ab\sin(bA)\sigma _{A}\right]^{2}}$	${\displaystyle \sigma _{f}\approx \left|ab\sin(bA)\sigma _{A}\right|}$
${\displaystyle f=a\tan \left(bA\right)\,}$	${\displaystyle \sigma _{f}^{2}\approx \left[ab\sec ^{2}(bA)\sigma _{A}\right]^{2}}$	${\displaystyle \sigma _{f}\approx \left|ab\sec ^{2}(bA)\sigma _{A}\right|}$
${\displaystyle f=A^{B}\,}$	${\displaystyle \sigma _{f}^{2}\approx f^{2}\left[\left({\frac {B}{A}}\sigma _{A}\right)^{2}+\left(\ln(A)\sigma _{B}\right)^{2}+2{\frac {B\ln(A)}{A}}\sigma _{AB}\right]}$	${\displaystyle \sigma _{f}\approx \left|f\right|{\sqrt {\left({\frac {B}{A}}\sigma _{A}\right)^{2}+\left(\ln(A)\sigma _{B}\right)^{2}+2{\frac {B\ln(A)}{A}}\sigma _{AB}}}}$
${\displaystyle f={\sqrt {aA^{2}\pm bB^{2}}}\,}$	${\displaystyle \sigma _{f}^{2}\approx \left({\frac {A}{f}}\right)^{2}a^{2}\sigma _{A}^{2}+\left({\frac {B}{f}}\right)^{2}b^{2}\sigma _{B}^{2}\pm 2ab{\frac {AB}{f^{2}}}\,\sigma _{AB}}$	${\displaystyle \sigma _{f}\approx {\sqrt {\left({\frac {A}{f}}\right)^{2}a^{2}\sigma _{A}^{2}+\left({\frac {B}{f}}\right)^{2}b^{2}\sigma _{B}^{2}\pm 2ab{\frac {AB}{f^{2}}}\,\sigma _{AB}}}}$

Close

For uncorrelated variables (${\displaystyle \rho _{AB}=0}$, ${\displaystyle \sigma _{AB}=0}$) expressions for more complicated functions can be derived by combining simpler functions. For example, repeated multiplication, assuming no correlation, gives

$${\displaystyle f=ABC;\qquad \left({\frac {\sigma _{f}}{f}}\right)^{2}\approx \left({\frac {\sigma _{A}}{A}}\right)^{2}+\left({\frac {\sigma _{B}}{B}}\right)^{2}+\left({\frac {\sigma _{C}}{C}}\right)^{2}.}$$

For the case ${\displaystyle f=AB}$ we also have Goodman's expression^[7] for the exact variance: for the uncorrelated case it is

$${\displaystyle V(XY)=E(X)^{2}V(Y)+E(Y)^{2}V(X)+E((X-E(X))^{2}(Y-E(Y))^{2})}$$

and therefore we have:

$${\displaystyle \sigma _{f}^{2}=A^{2}\sigma _{B}^{2}+B^{2}\sigma _{A}^{2}+\sigma _{A}^{2}\sigma _{B}^{2}}$$