Uncertainty¶

Types of Uncertainty¶

Others’ knowledge Our knowledge	Known	Unknown
Known	Things we are certain of	We know there are things we can’t predict eg: Random Process
Unknown	Others know but you don’t know eg: Insufficient data	Completely unexpected/unforeseeable events eg: Unknown distribution

	Aleatoric	Epistemic
Uncertainty in	Data	Model
Cause	Noisy input data Measurement errors	Missing training data
Describes confidence in	Input data	Prediction
Reducible through more training data	❌	✅
Can be learnt by model	✅	❌
Solution	Better instruments/measurements	Get more data

Uncertainty Intervals¶

You can obtain uncertainty using

	Concept	Limitations
Asymptotic approach	Central limit theorem	- Requires large sample size to satisfy asymptotic condition - Assumes normal distribution of errors - Assumes homoscedascity - Requires appropriate formula for calculating standard error (not possible for complex models)
Bootstrapping (preferred)	Random sampling with replacement	Higher computation cost

\[ \hat y \pm t_{\alpha/2} \times \text{SE} \]

	Coefficient Confidence Interval	Response Confidence Interval	Response Prediction Interval
Denotation	$\sigma_{\hat \beta}$	$\sigma\Big(\hat \mu(y \vert x) \Big)$	$\sigma\Big( \hat y \vert x \Big)$
The upper and lower bound for estimated __ at a given level of significance	$\hat \beta$	$\hat \mu(y \vert x)$	$\hat y \vert x$
SE (Standard Error) for Univariate Regression (Asymptotic Approach)	$\dfrac{\text{RMSE}}{\sqrt{\sum (x_{\text{pred}_\text{cent}} )^2}}$	$\text{RMSE} \times \sqrt{ \dfrac{1}{n} + \dfrac{(x_{\text{pred}_\text{cent}} )^2}{n \sigma_x^2}}$	$\text{RMSE} \times \sqrt{ \textcolor{hotpink}{1 +} \dfrac{1}{n} + \dfrac{(x_{\text{pred}_\text{cent}})^2}{n \sigma_x^2}}$
SE (Standard Error) for Multivariate Regression (Asymptotic Approach)	${\text{CovMatrix}_\beta}_{ii}$	$\text{RMSE} \times \sqrt{\dfrac{1}{n} + J'_{{x_\text{pred}}_\text{cent}} \ \text{CovMatrix}_{X} \ J'_{{x_\text{pred}}_\text{cent}}}$	$\text{RMSE} \times \sqrt{\textcolor{hotpink}{1+} \dfrac{1}{n} + x'_{\text{pred}_\text{cent}} \ \text{CovMatrix}_{X} \ x_{\text{pred}_\text{cent}}}$

\[ \begin{aligned} x'_{\text{pred}_\text{cent}} &= x_\text{pred} - \bar X \\ X_\text{cent} &= X - \bar X \\ \\ \text{CovMatrix}_{\beta} &= \text{RMSE} \times \sqrt{\text{CovMatrix}_{X}} \\ \text{CovMatrix}_{X} &= H^{-1} \\ H^{-1} &\approx (J' J)^{-1} \\ J &= X \text{ for degree 1} \end{aligned} \]

Where

$J$: Jacobean matrix
$H$: Hessian matrix

High values for non-diagonal elements of $\text{Cov}_\beta$ means that the errors of $\beta$ are correlated with each other.

Degree of freedom $= n - k - 1$, where

$n =$ sample size
$k=$ no of input variables

Confidence and prediction intervals are narrowest at $X = \bar X$, and get wider further from this point.

Under homoskedasticity, $$ \begin{aligned} \hat V(\hat \beta) &= (X' X)^{-1} \hat \sigma^2 \ &=\dfrac{\hat \sigma^2}{\hat u_j' \hat u_j} \end{aligned} $$

Note¶

RMSE = RMSE of validation data
If your validation error distribution is not normal, or you have a lot of data, you can use the quantiles of validation error distribution for the confidence intervals

Intervals using Models’ Prediction¶

For each data point, take __ of multiple models

average
5^th quantile
95^th quantile

Predictive Density¶

Describes the full probabilistic distribution $\forall x$

Trajectories/Scena rios¶

Equally-likely samples of multivariate predictive densities

Uncertainty Propagation¶

Function	Variance
$aA$	$= a^2\sigma_A^2$
$aA + bB$	$= a^2\sigma_A^2 + b^2\sigma_B^2 + 2ab\,\text{Cov(A, B)}$
$aA - bB$	$= a^2\sigma_A^2 + b^2\sigma_B^2 - 2ab\,\text{Cov(A, B)}$
$AB$	$\approx f^2 \left[\left(\frac{\sigma_A}{A}\right)^2 + \left(\frac{\sigma_B}{B}\right)^2 + 2\frac{\text{Cov(A, B)}}{AB} \right]$
$\frac{A}{B}$	$\approx f^2 \left[\left(\frac{\sigma_A}{A}\right)^2 + \left(\frac{\sigma_B}{B}\right)^2 - 2\frac{\text{Cov(A, B)}}{AB} \right]$
$\frac{A}{A+B}$	$\approx \frac{f^2}{\left(A+B\right)^2} \left(\frac{B^2}{A^2}\sigma_A^2 +\sigma_B^2 - 2\frac{B}{A} \text{Cov(A, B)} \right)$
$a A^{b}$	$\approx \left( {a}{b}{A}^{b-1}{\sigma_A} \right)^2 = \left( \frac{{f}{b}{\sigma_A}}{A} \right)^2$
$a \ln(bA)$	$\approx \left(a \frac{\sigma_A}{A} \right)^2$[^4]
$a \log_{10}(bA)$	$\approx \left(a \frac{\sigma_A}{A \ln(10)} \right)^2$[^5]
$a e^{bA}$	$\approx f^2 \left( b\sigma_A \right)^2$[^6]
$a^{bA}$	$\approx f^2 (b\ln(a)\sigma_A)^2$
$a \sin(bA)$	$\approx \left[ a b \cos(b A) \sigma_A \right]^2$
$a \cos \left( b A \right)$	$\approx \left[ a b \sin(b A) \sigma_A \right]^2$
$a \tan \left( b A \right)$	$\left[ a b \sec^2(b A) \sigma_A \right]^2$
$A^B$	$\approx f^2 \left[ \left( \frac{B}{A}\sigma_A \right)^2 +\left( \ln(A)\sigma_B \right)^2 + 2 \frac{B \ln(A)}{A} \text{Cov(A, B)} \right]$
$\sqrt{aA^2 \pm bB^2}$	$\approx \left(\frac{A}{f}\right)^2 a^2\sigma_A^2 + \left(\frac{B}{f}\right)^2 b^2\sigma_B^2 \pm 2ab\frac{AB}{f^2}\,\text{Cov(A, B)}$

For uncorrelated variables ($\rho_{AB}=0$, $\text{Cov(A, B)}=0$) expressions for more complicated functions can be derived by combining simpler functions. For example, repeated multiplication, assuming no correlation, gives $f = ABC; \qquad \left(\frac{\sigma_f}{f}\right)^2 \approx \left(\frac{\sigma_A}{A}\right)^2 + \left(\frac{\sigma_B}{B}\right)^2+ \left(\frac{\sigma_C}{C}\right)^2.$

For the case $f = AB$ we also have Goodman's expression[^7] for the exact variance: for the uncorrelated case it is $V(XY)= E(X)^2 V(Y) + E(Y)^2 V(X) + E((X-E(X))^2 (Y-E(Y))^2)$ and therefore we have: $\sigma_f^2 = A^2\sigma_B^2 + B^2\sigma_A^2 + \sigma_A^2\sigma_B^2$

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 ($\rho_{AB}\to 1$) 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 ($\rho_{AB}\to -1$) will further increase the variance of the difference, compared to the uncorrelated case.

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

Value at Risk Models¶

Derive the risk profile of the firm
Protect firm against unacceptably large concentrations
Quantify potential losses

Collect data
Graph the data to inspect data quality
Transform prices data into returns form (percentage diff of prices)
Look at the frequency distribution
Obtain the standard deviation (volatility)
Multiply volatility with one-sided $Z_1$ to estimate 99% worst-case loss

Classification¶

\[ [\text{Bin}(n, p)_{1-\alpha/2}, \text{Bin}(n, p)_{\alpha/2}] \]

Last Updated: 2024-12-26 ; Contributors: AhmedThahir, web-flow

	Coefficient Confidence Interval	Response Confidence Interval	Response Prediction Interval
Denotation	\(\sigma_{\hat \beta}\)	\(\sigma\Big(\hat \mu(y \vert x) \Big)\)	\(\sigma\Big( \hat y \vert x \Big)\)
The upper and lower bound for estimated __ at a given level of significance	\(\hat \beta\)	\(\hat \mu(y \vert x)\)	\(\hat y \vert x\)
SE (Standard Error) for Univariate Regression (Asymptotic Approach)	\(\dfrac{\text{RMSE}}{\sqrt{\sum (x_{\text{pred}_\text{cent}} )^2}}\)	\(\text{RMSE} \times \sqrt{ \dfrac{1}{n} + \dfrac{(x_{\text{pred}_\text{cent}} )^2}{n \sigma_x^2}}\)	\(\text{RMSE} \times \sqrt{ \textcolor{hotpink}{1 +} \dfrac{1}{n} + \dfrac{(x_{\text{pred}_\text{cent}})^2}{n \sigma_x^2}}\)
SE (Standard Error) for Multivariate Regression (Asymptotic Approach)	\({\text{CovMatrix}_\beta}_{ii}\)	\(\text{RMSE} \times \sqrt{\dfrac{1}{n} + J'_{{x_\text{pred}}_\text{cent}} \ \text{CovMatrix}_{X} \ J'_{{x_\text{pred}}_\text{cent}}}\)	\(\text{RMSE} \times \sqrt{\textcolor{hotpink}{1+} \dfrac{1}{n} + x'_{\text{pred}_\text{cent}} \ \text{CovMatrix}_{X} \ x_{\text{pred}_\text{cent}}}\)

Function	Variance
\(aA\)	\(= a^2\sigma_A^2\)
\(aA + bB\)	\(= a^2\sigma_A^2 + b^2\sigma_B^2 + 2ab\,\text{Cov(A, B)}\)
\(aA - bB\)	\(= a^2\sigma_A^2 + b^2\sigma_B^2 - 2ab\,\text{Cov(A, B)}\)
\(AB\)	\(\approx f^2 \left[\left(\frac{\sigma_A}{A}\right)^2 + \left(\frac{\sigma_B}{B}\right)^2 + 2\frac{\text{Cov(A, B)}}{AB} \right]\)
\(\frac{A}{B}\)	\(\approx f^2 \left[\left(\frac{\sigma_A}{A}\right)^2 + \left(\frac{\sigma_B}{B}\right)^2 - 2\frac{\text{Cov(A, B)}}{AB} \right]\)
\(\frac{A}{A+B}\)	\(\approx \frac{f^2}{\left(A+B\right)^2} \left(\frac{B^2}{A^2}\sigma_A^2 +\sigma_B^2 - 2\frac{B}{A} \text{Cov(A, B)} \right)\)
\(a A^{b}\)	\(\approx \left( {a}{b}{A}^{b-1}{\sigma_A} \right)^2 = \left( \frac{{f}{b}{\sigma_A}}{A} \right)^2\)
\(a \ln(bA)\)	\(\approx \left(a \frac{\sigma_A}{A} \right)^2\)[^4]
\(a \log_{10}(bA)\)	\(\approx \left(a \frac{\sigma_A}{A \ln(10)} \right)^2\)[^5]
\(a e^{bA}\)	\(\approx f^2 \left( b\sigma_A \right)^2\)[^6]
\(a^{bA}\)	\(\approx f^2 (b\ln(a)\sigma_A)^2\)
\(a \sin(bA)\)	\(\approx \left[ a b \cos(b A) \sigma_A \right]^2\)
\(a \cos \left( b A \right)\)	\(\approx \left[ a b \sin(b A) \sigma_A \right]^2\)
\(a \tan \left( b A \right)\)	\(\left[ a b \sec^2(b A) \sigma_A \right]^2\)
\(A^B\)	\(\approx f^2 \left[ \left( \frac{B}{A}\sigma_A \right)^2 +\left( \ln(A)\sigma_B \right)^2 + 2 \frac{B \ln(A)}{A} \text{Cov(A, B)} \right]\)
\(\sqrt{aA^2 \pm bB^2}\)	\(\approx \left(\frac{A}{f}\right)^2 a^2\sigma_A^2 + \left(\frac{B}{f}\right)^2 b^2\sigma_B^2 \pm 2ab\frac{AB}{f^2}\,\text{Cov(A, B)}\)