STA457 Notation Index

Standard notation used across all STA457 notes.

Time Series

SymbolMeaningNotes
$\{X_t\}$ or $\{X_t\}_{t \in T}$Stochastic processCapital = random variable
$x_t$Observed value at time $t$Lowercase = realization
$n$Number of observations
$h$Time lag

Components and Decomposition

SymbolMeaningNotes
$m_t$Deterministic trendClassical Decomposition
$s_t$Seasonal componentPeriod $d$: $s_t = s_{t+d}$
$Y_t$Random (noise) componentZero-mean

Operators

SymbolMeaningDefinition
$B$Backshift Operator$BX_t = X_{t-1}$
$B^k$$k$-step backshift$B^k X_t = X_{t-k}$
$\nabla$Difference operator$\nabla = 1 - B$
$\nabla^p$$p$-th order difference$(1-B)^p$
$\nabla_d$Lag-$d$ difference$1 - B^d$
$\phi(B)$AR polynomial in $B$$1 - \phi_1 B - \cdots - \phi_p B^p$
$\theta(B)$MA polynomial in $B$$1 + \theta_1 B + \cdots + \theta_q B^q$

Statistical Functions

SymbolMeaningNotes
$\mu_X$ or $\mu$Mean of $\{X_t\}$Constant under Stationarity
$\sigma_X^2$ or $\sigma^2$Variance of $\{X_t\}$$= \gamma_X(0)$
$\gamma_X(h)$ACVF at lag $h$$\text{Cov}(X_t, X_{t+h})$
$\rho_X(h)$ACF at lag $h$$\gamma_X(h)/\gamma_X(0)$
$\hat{\gamma}(h)$Sample ACVF$n^{-1}\sum(x_t - \bar{x})(x_{t+h}-\bar{x})$
$\hat{\rho}(h)$Sample ACF$\hat{\gamma}(h)/\hat{\gamma}(0)$

Model Parameters

SymbolMeaningContext
$\phi_1, \dots, \phi_p$AR coefficientsAutoregressive Process
$\theta_1, \dots, \theta_q$MA coefficientsMoving Average Process
$\sigma^2$White noise variance$\{Z_t\} \sim \text{WN}(0, \sigma^2)$
$\{Z_t\}$White noise processInput/innovation
$\{a_j\}$Filter coefficientsLinear Filter

Prediction

SymbolMeaningNotes
$P_n X_{n+h}$Best Linear Predictor of $X_{n+h}$ given $X_1, \dots, X_n$Minimizes MSE
$\text{MSE}$Mean squared error$\mathbb{E}[(X_{n+h} - P_n X_{n+h})^2]$
$\Gamma_n$Covariance matrix$(\gamma_X(i-j))_{i,j=1}^n$

Common Abbreviations

AbbreviationMeaning
WNWhite Noise Process
AR($p$)Autoregressive of order $p$
MA($q$)Moving average of order $q$
ARMA($p,q$)Autoregressive moving average
ACVFAutocovariance function
ACFAutocorrelation function
BLPBest linear predictor