Propagation of uncertainty

On this page

Effect of variables' uncertainties (or errors, more specifically random errors) on the uncertainty of a function based on them

In statistics, propagation of uncertainty refers to how variables' uncertainties affect the uncertainty of a function based on them, especially when variables come from experimental measurements with inherent measurement limitations. Uncertainty can be expressed through absolute error, relative error, or most commonly the standard deviation. Characterizing uncertainty often involves specifying its probability distribution, enabling derivation of confidence limits. When uncertainties are correlated, covariance must be considered. For complex cases, techniques like the Monte Carlo method are used, sometimes assisted by surrogate models or parallel computing to manage computational cost.

We don't have any images related to Propagation of uncertainty yet.

You can add one yourself here.

We don't have any YouTube videos related to Propagation of uncertainty yet.

You can add one yourself here.

We don't have any PDF documents related to Propagation of uncertainty yet.

You can add one yourself here.

We don't have any Books related to Propagation of uncertainty yet.

You can add one yourself here.

We don't have any archived web articles related to Propagation of uncertainty yet.

You can submit a link to a page to archive here.

Linear combinations

Let { f k ( x 1 , x 2 , … , x n ) } {\displaystyle \{f_{k}(x_{1},x_{2},\dots ,x_{n})\}} be a set of m functions, which are linear combinations of n {\displaystyle n} variables x 1 , x 2 , … , x n {\displaystyle x_{1},x_{2},\dots ,x_{n}} with combination coefficients A k 1 , A k 2 , … , A k n , ( k = 1 , … , m ) {\displaystyle A_{k1},A_{k2},\dots ,A_{kn},(k=1,\dots ,m)} : f k = ∑ i = 1 n A k i x i , {\displaystyle f_{k}=\sum _{i=1}^{n}A_{ki}x_{i},} or in matrix notation, f = A x . {\displaystyle \mathbf {f} =\mathbf {A} \mathbf {x} .}

Also let the variance–covariance matrix of x = (x1, ..., xn) be denoted by Σ x {\displaystyle {\boldsymbol {\Sigma }}^{x}} and let the mean value be denoted by μ {\displaystyle {\boldsymbol {\mu }}} : Σ x = E [ ( x − μ ) ⊗ ( x − μ ) ] = ( σ 1 2 σ 12 σ 13 ⋯ σ 21 σ 2 2 σ 23 ⋯ σ 31 σ 32 σ 3 2 ⋯ ⋮ ⋮ ⋮ ⋱ ) = ( Σ 11 x Σ 12 x Σ 13 x ⋯ Σ 21 x Σ 22 x Σ 23 x ⋯ Σ 31 x Σ 32 x Σ 33 x ⋯ ⋮ ⋮ ⋮ ⋱ ) . {\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 Σ f {\displaystyle {\boldsymbol {\Sigma }}^{f}} of f is given by Σ f = E [ ( f − E [ f ] ) ⊗ ( f − E [ f ] ) ] = E [ A ( x − μ ) ⊗ A ( x − μ ) ] = A E [ ( x − μ ) ⊗ ( x − μ ) ] A T = A Σ x A T . {\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 Σ f = A Σ x A T {\displaystyle {\boldsymbol {\Sigma }}^{f}=\mathbf {A} {\boldsymbol {\Sigma }}^{x}\mathbf {A} ^{\mathrm {T} }} reads Σ i j f = ∑ k n ∑ l n A i k Σ k l x A j l . {\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 Σ i j f = ∑ k n A i k Σ k x A j k , {\displaystyle \Sigma _{ij}^{f}=\sum _{k}^{n}A_{ik}\Sigma _{k}^{x}A_{jk},} where Σ k x = σ x k 2 {\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 Σ x {\displaystyle {\boldsymbol {\Sigma }}^{x}} is a diagonal matrix, Σ f {\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): f = ∑ i n a i x i = a x , {\displaystyle f=\sum _{i}^{n}a_{i}x_{i}=\mathbf {ax} ,} σ f 2 = ∑ i n ∑ j n a i Σ i j x a j = a Σ x a T . {\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 σ i j {\displaystyle \sigma _{ij}} can be expressed in terms of the correlation coefficient ρ i j {\displaystyle \rho _{ij}} by σ i j = ρ i j σ i σ j {\displaystyle \sigma _{ij}=\rho _{ij}\sigma _{i}\sigma _{j}} , so that an alternative expression for the variance of f is σ f 2 = ∑ i n a i 2 σ i 2 + ∑ i n ∑ j ( j ≠ i ) n a i a j ρ i j σ i σ j . {\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 σ f 2 = ∑ i n a i 2 σ i 2 . {\displaystyle \sigma _{f}^{2}=\sum _{i}^{n}a_{i}^{2}\sigma _{i}^{2}.}

In the simple case of identical coefficients and variances, we find σ f = n | a | σ . {\displaystyle \sigma _{f}={\sqrt {n}}\,|a|\sigma .}

For the arithmetic mean, a = 1 / n {\displaystyle a=1/n} , the result is the standard error of the mean: σ f = σ n . {\displaystyle \sigma _{f}={\frac {\sigma }{\sqrt {n}}}.}

Non-linear combinations

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: f k ≈ f k 0 + ∑ i n ∂ f k ∂ x i x i {\displaystyle f_{k}\approx f_{k}^{0}+\sum _{i}^{n}{\frac {\partial f_{k}}{\partial {x_{i}}}}x_{i}} where ∂ f k / ∂ x i {\displaystyle \partial f_{k}/\partial x_{i}} denotes the partial derivative of fk with respect to the i-th variable, evaluated at the mean value of all components of vector x. Or in matrix notation, f ≈ f 0 + J x {\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, Aki and Akj by the partial derivatives, ∂ f k ∂ x i {\displaystyle {\frac {\partial f_{k}}{\partial x_{i}}}} and ∂ f k ∂ x j {\displaystyle {\frac {\partial f_{k}}{\partial x_{j}}}} . In matrix notation,⁸ Σ f = J Σ x J ⊤ . {\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 J = A {\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 s f = ( ∂ f ∂ x ) 2 s x 2 + ( ∂ f ∂ y ) 2 s y 2 + ( ∂ f ∂ z ) 2 s z 2 + ⋯ {\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 s f {\displaystyle s_{f}} represents the standard deviation of the function f {\displaystyle f} , s x {\displaystyle s_{x}} represents the standard deviation of x {\displaystyle x} , s y {\displaystyle s_{y}} represents the standard deviation of y {\displaystyle y} , and so forth.

This formula is based on the linear characteristics of the gradient of f {\displaystyle f} and therefore it is a good estimation for the standard deviation of f {\displaystyle f} as long as s x , s y , s z , … {\displaystyle s_{x},s_{y},s_{z},\ldots } are small enough. Specifically, the linear approximation of f {\displaystyle f} has to be close to f {\displaystyle f} inside a neighbourhood of radius s x , s y , s z , … {\displaystyle s_{x},s_{y},s_{z},\ldots } .¹⁰

Example

Any non-linear differentiable function, f ( a , b ) {\displaystyle f(a,b)} , of two variables, a {\displaystyle a} and b {\displaystyle b} , can be expanded as f ≈ f 0 + ∂ f ∂ a a + ∂ f ∂ b b . {\displaystyle f\approx f^{0}+{\frac {\partial f}{\partial a}}a+{\frac {\partial f}{\partial b}}b.} If we take the variance on both sides and use the formula¹¹ for the variance of a linear combination of variables Var ⁡ ( a X + b Y ) = a 2 Var ⁡ ( X ) + b 2 Var ⁡ ( Y ) + 2 a b Cov ⁡ ( X , Y ) , {\displaystyle \operatorname {Var} (aX+bY)=a^{2}\operatorname {Var} (X)+b^{2}\operatorname {Var} (Y)+2ab\operatorname {Cov} (X,Y),} then we obtain σ f 2 ≈ | ∂ f ∂ a | 2 σ a 2 + | ∂ f ∂ b | 2 σ b 2 + 2 ∂ f ∂ a ∂ f ∂ b σ a b , {\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 σ f {\displaystyle \sigma _{f}} is the standard deviation of the function f {\displaystyle f} , σ a {\displaystyle \sigma _{a}} is the standard deviation of a {\displaystyle a} , σ b {\displaystyle \sigma _{b}} is the standard deviation of b {\displaystyle b} and σ a b = σ a σ b ρ a b {\displaystyle \sigma _{ab}=\sigma _{a}\sigma _{b}\rho _{ab}} is the covariance between a {\displaystyle a} and b {\displaystyle b} .

In the particular case that f = a b {\displaystyle f=ab} , ∂ f ∂ a = b {\displaystyle {\frac {\partial f}{\partial a}}=b} , ∂ f ∂ b = a {\displaystyle {\frac {\partial f}{\partial b}}=a} . Then σ f 2 ≈ b 2 σ a 2 + a 2 σ b 2 + 2 a b σ a b {\displaystyle \sigma _{f}^{2}\approx b^{2}\sigma _{a}^{2}+a^{2}\sigma _{b}^{2}+2ab\,\sigma _{ab}} or ( σ f f ) 2 ≈ ( σ a a ) 2 + ( σ b b ) 2 + 2 ( σ a a ) ( σ b b ) ρ a b {\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 ρ a b {\displaystyle \rho _{ab}} is the correlation between a {\displaystyle a} and b {\displaystyle b} .

When the variables a {\displaystyle a} and b {\displaystyle b} are uncorrelated, ρ a b = 0 {\displaystyle \rho _{ab}=0} . Then ( σ f f ) 2 ≈ ( σ a a ) 2 + ( σ b b ) 2 . {\displaystyle \left({\frac {\sigma _{f}}{f}}\right)^{2}\approx \left({\frac {\sigma _{a}}{a}}\right)^{2}+\left({\frac {\sigma _{b}}{b}}\right)^{2}.}

Caveats and warnings

Error estimates for non-linear functions are biased on account of using a truncated series expansion. The extent of this bias depends on the nature of the function. For example, the bias on the error calculated for log(1+x) increases as x increases, since the expansion to x is a good approximation only when x is near zero.

For highly non-linear functions, there exist five categories of probabilistic approaches for uncertainty propagation;¹² see Uncertainty quantification for details.

Reciprocal and shifted reciprocal

Main article: Reciprocal normal distribution

In the special case of the inverse or reciprocal 1 / B {\displaystyle 1/B} , where B = N ( 0 , 1 ) {\displaystyle B=N(0,1)} follows a standard normal distribution, the resulting distribution is a reciprocal standard normal distribution, and there is no definable variance.¹³

However, in the slightly more general case of a shifted reciprocal function 1 / ( p − B ) {\displaystyle 1/(p-B)} for B = N ( μ , σ ) {\displaystyle B=N(\mu ,\sigma )} following a general normal distribution, then mean and variance statistics do exist in a principal value sense, if the difference between the pole p {\displaystyle p} and the mean μ {\displaystyle \mu } is real-valued.¹⁴

Ratios

Main article: Normal ratio distribution

Ratios are also problematic; normal approximations exist under certain conditions.

Example formulae

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

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

Function	Variance	Standard deviation
f = a A {\displaystyle f=aA\,}	σ f 2 = a 2 σ A 2 {\displaystyle \sigma _{f}^{2}=a^{2}\sigma _{A}^{2}}	σ f = \| a \| σ A {\displaystyle \sigma _{f}=\|a\|\sigma _{A}}
f = A + B {\displaystyle f=A+B}	σ f 2 = σ A 2 + σ B 2 + 2 σ A B {\displaystyle \sigma _{f}^{2}=\sigma _{A}^{2}+\sigma _{B}^{2}+2\sigma _{AB}}	σ f = σ A 2 + σ B 2 + 2 σ A B {\displaystyle \sigma _{f}={\sqrt {\sigma _{A}^{2}+\sigma _{B}^{2}+2\sigma _{AB}}}}
f = A − B {\displaystyle f=A-B}	σ f 2 = σ A 2 + σ B 2 − 2 σ A B {\displaystyle \sigma _{f}^{2}=\sigma _{A}^{2}+\sigma _{B}^{2}-2\sigma _{AB}}	σ f = σ A 2 + σ B 2 − 2 σ A B {\displaystyle \sigma _{f}={\sqrt {\sigma _{A}^{2}+\sigma _{B}^{2}-2\sigma _{AB}}}}
f = a A + b B {\displaystyle f=aA+bB}	σ f 2 = a 2 σ A 2 + b 2 σ B 2 + 2 a b σ A B {\displaystyle \sigma _{f}^{2}=a^{2}\sigma _{A}^{2}+b^{2}\sigma _{B}^{2}+2ab\,\sigma _{AB}}	σ f = a 2 σ A 2 + b 2 σ B 2 + 2 a b σ A B {\displaystyle \sigma _{f}={\sqrt {a^{2}\sigma _{A}^{2}+b^{2}\sigma _{B}^{2}+2ab\,\sigma _{AB}}}}
f = a A − b B {\displaystyle f=aA-bB}	σ f 2 = a 2 σ A 2 + b 2 σ B 2 − 2 a b σ A B {\displaystyle \sigma _{f}^{2}=a^{2}\sigma _{A}^{2}+b^{2}\sigma _{B}^{2}-2ab\,\sigma _{AB}}	σ f = a 2 σ A 2 + b 2 σ B 2 − 2 a b σ A B {\displaystyle \sigma _{f}={\sqrt {a^{2}\sigma _{A}^{2}+b^{2}\sigma _{B}^{2}-2ab\,\sigma _{AB}}}}
f = A B {\displaystyle f=AB}	σ f 2 ≈ f 2 [ ( σ A A ) 2 + ( σ B B ) 2 + 2 σ A B 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]} ¹⁵ ¹⁶	σ f ≈ \| f \| ( σ A A ) 2 + ( σ B B ) 2 + 2 σ A B A B {\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}}}}}
f = A B {\displaystyle f={\frac {A}{B}}}	σ f 2 ≈ f 2 [ ( σ A A ) 2 + ( σ B B ) 2 − 2 σ A B 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	σ f ≈ \| f \| ( σ A A ) 2 + ( σ B B ) 2 − 2 σ A B A B {\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}}}}}
f = A A + B {\displaystyle f={\frac {A}{A+B}}}	σ f 2 ≈ f 2 ( A + B ) 2 ( B 2 A 2 σ A 2 + σ B 2 − 2 B A σ 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)}	σ f ≈ \| f A + B \| B 2 A 2 σ A 2 + σ B 2 − 2 B A σ A B {\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}}}}
f = a A b {\displaystyle f=aA^{b}}	σ f 2 ≈ ( a b A b − 1 σ A ) 2 = ( f b σ A A ) 2 {\displaystyle \sigma _{f}^{2}\approx \left({a}{b}{A}^{b-1}{\sigma _{A}}\right)^{2}=\left({\frac {{f}{b}{\sigma _{A}}}{A}}\right)^{2}}	σ f ≈ \| a b A b − 1 σ A \| = \| f b σ A A \| {\displaystyle \sigma _{f}\approx \left\|{a}{b}{A}^{b-1}{\sigma _{A}}\right\|=\left\|{\frac {{f}{b}{\sigma _{A}}}{A}}\right\|}
f = a ln ⁡ ( b A ) {\displaystyle f=a\ln(bA)}	σ f 2 ≈ ( a σ A A ) 2 {\displaystyle \sigma _{f}^{2}\approx \left(a{\frac {\sigma _{A}}{A}}\right)^{2}} ¹⁸	σ f ≈ \| a σ A A \| {\displaystyle \sigma _{f}\approx \left\|a{\frac {\sigma _{A}}{A}}\right\|}
f = a log 10 ⁡ ( b A ) {\displaystyle f=a\log _{10}(bA)}	σ f 2 ≈ ( a σ A A ln ⁡ ( 10 ) ) 2 {\displaystyle \sigma _{f}^{2}\approx \left(a{\frac {\sigma _{A}}{A\ln(10)}}\right)^{2}} ¹⁹	σ f ≈ \| a σ A A ln ⁡ ( 10 ) \| {\displaystyle \sigma _{f}\approx \left\|a{\frac {\sigma _{A}}{A\ln(10)}}\right\|}
f = a e b A {\displaystyle f=ae^{bA}}	σ f 2 ≈ f 2 ( b σ A ) 2 {\displaystyle \sigma _{f}^{2}\approx f^{2}\left(b\sigma _{A}\right)^{2}} ²⁰	σ f ≈ \| f \| \| ( b σ A ) \| {\displaystyle \sigma _{f}\approx \left\|f\right\|\left\|\left(b\sigma _{A}\right)\right\|}
f = a b A {\displaystyle f=a^{bA}}	σ f 2 ≈ f 2 ( b ln ⁡ ( a ) σ A ) 2 {\displaystyle \sigma _{f}^{2}\approx f^{2}(b\ln(a)\sigma _{A})^{2}}	σ f ≈ \| f \| \| b ln ⁡ ( a ) σ A \| {\displaystyle \sigma _{f}\approx \left\|f\right\|\left\|b\ln(a)\sigma _{A}\right\|}
f = a sin ⁡ ( b A ) {\displaystyle f=a\sin(bA)}	σ f 2 ≈ [ a b cos ⁡ ( b A ) σ A ] 2 {\displaystyle \sigma _{f}^{2}\approx \left[ab\cos(bA)\sigma _{A}\right]^{2}}	σ f ≈ \| a b cos ⁡ ( b A ) σ A \| {\displaystyle \sigma _{f}\approx \left\|ab\cos(bA)\sigma _{A}\right\|}
f = a cos ⁡ ( b A ) {\displaystyle f=a\cos \left(bA\right)\,}	σ f 2 ≈ [ a b sin ⁡ ( b A ) σ A ] 2 {\displaystyle \sigma _{f}^{2}\approx \left[ab\sin(bA)\sigma _{A}\right]^{2}}	σ f ≈ \| a b sin ⁡ ( b A ) σ A \| {\displaystyle \sigma _{f}\approx \left\|ab\sin(bA)\sigma _{A}\right\|}
f = a tan ⁡ ( b A ) {\displaystyle f=a\tan \left(bA\right)\,}	σ f 2 ≈ [ a b sec 2 ⁡ ( b A ) σ A ] 2 {\displaystyle \sigma _{f}^{2}\approx \left[ab\sec ^{2}(bA)\sigma _{A}\right]^{2}}	σ f ≈ \| a b sec 2 ⁡ ( b A ) σ A \| {\displaystyle \sigma _{f}\approx \left\|ab\sec ^{2}(bA)\sigma _{A}\right\|}
f = A B {\displaystyle f=A^{B}}	σ f 2 ≈ f 2 [ ( B A σ A ) 2 + ( ln ⁡ ( A ) σ B ) 2 + 2 B ln ⁡ ( A ) A σ 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]}	σ f ≈ \| f \| ( B A σ A ) 2 + ( ln ⁡ ( A ) σ B ) 2 + 2 B ln ⁡ ( A ) A σ A B {\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}}}}
f = a A 2 ± b B 2 {\displaystyle f={\sqrt {aA^{2}\pm bB^{2}}}}	σ f 2 ≈ ( A f ) 2 a 2 σ A 2 + ( B f ) 2 b 2 σ B 2 ± 2 a b A B f 2 σ A B {\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}}	σ f ≈ ( A f ) 2 a 2 σ A 2 + ( B f ) 2 b 2 σ B 2 ± 2 a b A B f 2 σ A B {\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}}}}

For uncorrelated variables ( ρ A B = 0 {\displaystyle \rho _{AB}=0} , σ A B = 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 f = A B C ; ( σ f f ) 2 ≈ ( σ A A ) 2 + ( σ B B ) 2 + ( σ C C ) 2 . {\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 f = A B {\displaystyle f=AB} we also have Goodman's expression²¹ for the exact variance: for the uncorrelated case it is V ( X Y ) = E ( X ) 2 V ( Y ) + E ( Y ) 2 V ( X ) + E ( ( X − E ( X ) ) 2 ( Y − E ( Y ) ) 2 ) , {\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 σ f 2 = A 2 σ B 2 + B 2 σ A 2 + σ A 2 σ B 2 . {\displaystyle \sigma _{f}^{2}=A^{2}\sigma _{B}^{2}+B^{2}\sigma _{A}^{2}+\sigma _{A}^{2}\sigma _{B}^{2}.}

Effect of correlation on differences

If A and B are uncorrelated, their difference A − B will have more variance than either of them. An increasing positive correlation ( ρ A B → 1 {\displaystyle \rho _{AB}\to 1} ) will decrease the variance of the difference, converging to zero variance for perfectly correlated variables with the same variance. On the other hand, a negative correlation ( ρ A B → − 1 {\displaystyle \rho _{AB}\to -1} ) will further increase the variance of the difference, compared to the uncorrelated case.

For example, the self-subtraction f = A − A has zero variance σ f 2 = 0 {\displaystyle \sigma _{f}^{2}=0} only if the variate is perfectly autocorrelated ( ρ A = 1 {\displaystyle \rho _{A}=1} ). If A is uncorrelated, ρ A = 0 , {\displaystyle \rho _{A}=0,} then the output variance is twice the input variance, σ f 2 = 2 σ A 2 . {\displaystyle \sigma _{f}^{2}=2\sigma _{A}^{2}.} And if A is perfectly anticorrelated, ρ A = − 1 , {\displaystyle \rho _{A}=-1,} then the input variance is quadrupled in the output, σ f 2 = 4 σ A 2 {\displaystyle \sigma _{f}^{2}=4\sigma _{A}^{2}} (notice 1 − ρ A = 2 {\displaystyle 1-\rho _{A}=2} for f = aA − aA in the table above).

Example calculations

Inverse tangent function

We can calculate the uncertainty propagation for the inverse tangent function as an example of using partial derivatives to propagate error.

Define f ( x ) = arctan ⁡ ( x ) , {\displaystyle f(x)=\arctan(x),} where Δ x {\displaystyle \Delta _{x}} is the absolute uncertainty on our measurement of x. The derivative of f(x) with respect to x is d f d x = 1 1 + x 2 . {\displaystyle {\frac {df}{dx}}={\frac {1}{1+x^{2}}}.}

Therefore, our propagated uncertainty is Δ f ≈ Δ x 1 + x 2 , {\displaystyle \Delta _{f}\approx {\frac {\Delta _{x}}{1+x^{2}}},} where Δ f {\displaystyle \Delta _{f}} is the absolute propagated uncertainty.

Resistance measurement

A practical application is an experiment in which one measures current, I, and voltage, V, on a resistor in order to determine the resistance, R, using Ohm's law, R = V / I.

Given the measured variables with uncertainties, I ± σI and V ± σV, and neglecting their possible correlation, the uncertainty in the computed quantity, σR, is:

σ R ≈ σ V 2 ( 1 I ) 2 + σ I 2 ( − V I 2 ) 2 = R ( σ V V ) 2 + ( σ I I ) 2 . {\displaystyle \sigma _{R}\approx {\sqrt {\sigma _{V}^{2}\left({\frac {1}{I}}\right)^{2}+\sigma _{I}^{2}\left({\frac {-V}{I^{2}}}\right)^{2}}}=R{\sqrt {\left({\frac {\sigma _{V}}{V}}\right)^{2}+\left({\frac {\sigma _{I}}{I}}\right)^{2}}}.}

External links

A detailed discussion of measurements and the propagation of uncertainty explaining the benefits of using error propagation formulas and Monte Carlo simulations instead of simple significance arithmetic
GUM, Guide to the Expression of Uncertainty in Measurement
EPFL An Introduction to Error Propagation, Derivation, Meaning and Examples of Cy = Fx Cx Fx'
uncertainties package, a program/library for transparently performing calculations with uncertainties (and error correlations).
soerp package, a Python program/library for transparently performing *second-order* calculations with uncertainties (and error correlations).
Joint Committee for Guides in Metrology (2011). JCGM 102: Evaluation of Measurement Data - Supplement 2 to the "Guide to the Expression of Uncertainty in Measurement" - Extension to Any Number of Output Quantities (PDF) (Technical report). JCGM. Retrieved 13 February 2013.
Uncertainty Calculator Propagate uncertainty for any expression

References

Kirchner, James. "Data Analysis Toolkit #5: Uncertainty Analysis and Error Propagation" (PDF). Berkeley Seismology Laboratory. University of California. Retrieved 22 April 2016. http://seismo.berkeley.edu/~kirchner/eps_120/Toolkits/Toolkit_05.pdf ↩
Kroese, D. P.; Taimre, T.; Botev, Z. I. (2011). Handbook of Monte Carlo Methods. John Wiley & Sons. ↩
Ranftl, Sascha; von der Linden, Wolfgang (2021-11-13). "Bayesian Surrogate Analysis and Uncertainty Propagation". Physical Sciences Forum. 3 (1): 6. arXiv:2101.04038. doi:10.3390/psf2021003006. ISSN 2673-9984. https://doi.org/10.3390%2Fpsf2021003006 ↩
Atanassova, E.; Gurov, T.; Karaivanova, A.; Ivanovska, S.; Durchova, M.; Dimitrov, D. (2016). "On the parallelization approaches for Intel MIC architecture". AIP Conference Proceedings. 1773 (1): 070001. Bibcode:2016AIPC.1773g0001A. doi:10.1063/1.4964983. /wiki/Bibcode_(identifier) ↩
Cunha Jr, A.; Nasser, R.; Sampaio, R.; Lopes, H.; Breitman, K. (2014). "Uncertainty quantification through the Monte Carlo method in a cloud computing setting". Computer Physics Communications. 185 (5): 1355–1363. arXiv:2105.09512. Bibcode:2014CoPhC.185.1355C. doi:10.1016/j.cpc.2014.01.006. S2CID 32376269. /wiki/ArXiv_(identifier) ↩
Lin, Y.; Wang, F.; Liu, B. (2018). "Random number generators for large-scale parallel Monte Carlo simulations on FPGA". Journal of Computational Physics. 360: 93–103. Bibcode:2018JCoPh.360...93L. doi:10.1016/j.jcp.2018.01.029. /wiki/Bibcode_(identifier) ↩
Goodman, Leo (1960). "On the Exact Variance of Products". Journal of the American Statistical Association. 55 (292): 708–713. doi:10.2307/2281592. JSTOR 2281592. /wiki/Leo_Goodman ↩
Ochoa1, Benjamin; Belongie, Serge "Covariance Propagation for Guided Matching" Archived 2011-07-20 at the Wayback Machine http://vision.ucsd.edu/sites/default/files/ochoa06.pdf ↩
Ku, H. H. (October 1966). "Notes on the use of propagation of error formulas". Journal of Research of the National Bureau of Standards. 70C (4): 262. doi:10.6028/jres.070c.025. ISSN 0022-4316. Retrieved 3 October 2012. http://nistdigitalarchives.contentdm.oclc.org/cdm/compoundobject/collection/p16009coll6/id/99848/rec/1 ↩
Clifford, A. A. (1973). Multivariate error analysis: a handbook of error propagation and calculation in many-parameter systems. John Wiley & Sons. ISBN 978-0470160558.[page needed] 978-0470160558 ↩
Soch, Joram (2020-07-07). "Variance of the linear combination of two random variables". The Book of Statistical Proofs. Retrieved 2022-01-29. https://statproofbook.github.io/P/var-lincomb.html ↩
Lee, S. H.; Chen, W. (2009). "A comparative study of uncertainty propagation methods for black-box-type problems". Structural and Multidisciplinary Optimization. 37 (3): 239–253. doi:10.1007/s00158-008-0234-7. S2CID 119988015. /wiki/Doi_(identifier) ↩
Johnson, Norman L.; Kotz, Samuel; Balakrishnan, Narayanaswamy (1994). Continuous Univariate Distributions, Volume 1. Wiley. p. 171. ISBN 0-471-58495-9. 0-471-58495-9 ↩
Lecomte, Christophe (May 2013). "Exact statistics of systems with uncertainties: an analytical theory of rank-one stochastic dynamic systems". Journal of Sound and Vibration. 332 (11): 2750–2776. Bibcode:2013JSV...332.2750L. doi:10.1016/j.jsv.2012.12.009. /wiki/Bibcode_(identifier) ↩
"A Summary of Error Propagation" (PDF). p. 2. Archived from the original (PDF) on 2016-12-13. Retrieved 2016-04-04. https://web.archive.org/web/20161213135602/http://ipl.physics.harvard.edu/wp-uploads/2013/03/PS3_Error_Propagation_sp13.pdf ↩
"Propagation of Uncertainty through Mathematical Operations" (PDF). p. 5. Retrieved 2016-04-04. http://web.mit.edu/fluids-modules/www/exper_techniques/2.Propagation_of_Uncertaint.pdf ↩
"Strategies for Variance Estimation" (PDF). p. 37. Retrieved 2013-01-18. http://www.sagepub.com/upm-data/6427_Chapter_4__Lee_%28Analyzing%29_I_PDF_6.pdf ↩
Harris, Daniel C. (2003), Quantitative chemical analysis (6th ed.), Macmillan, p. 56, ISBN 978-0-7167-4464-1 978-0-7167-4464-1 ↩
Harris, Daniel C. (2003), Quantitative chemical analysis (6th ed.), Macmillan, p. 56, ISBN 978-0-7167-4464-1 978-0-7167-4464-1 ↩
"Error Propagation tutorial" (PDF). Foothill College. October 9, 2009. Retrieved 2012-03-01. http://www.foothill.edu/psme/daley/tutorials_files/10.%20Error%20Propagation.pdf ↩
Goodman, Leo (1960). "On the Exact Variance of Products". Journal of the American Statistical Association. 55 (292): 708–713. doi:10.2307/2281592. JSTOR 2281592. /wiki/Leo_Goodman ↩

Linear combinations

Non-linear combinations

Simplification

Example

Caveats and warnings

Reciprocal and shifted reciprocal

Ratios

Example formulae

Effect of correlation on differences

Example calculations

Inverse tangent function

Resistance measurement

See also

Further reading

External links

References