MA(1) process
Summary
- A stochastic process \(\{ Y_t \}\) is called an moving average process of order 1 or an MA(1) process if
\[Y_t=μ+ε_t+αε_{t-1}\]
- where \(μ\) and \(α\) are parameters and \(\{ ε_t \}\) is white noise.
- An MA(1) process is always stationarity .
- Result: The expected value of an MA(1) process \(Y_t=μ+ε_t+αε_{t-1}\) :
\[E\left( Y_t \right)=μ\]
- If \(Y_t\) follows an MA(1) process, \(Y_t=μ+ε_t+αε_{t-1}\) , then \(y_t=Y_t-E\left( Y_t \right)\) will follow an MA(1) process with intercept 0
\[y_t=ε_t+αε_{t-1}\]
- This is notationally much more convenient. We have
\[E\left( y_t \right)=0\]
- Result: The variance of an MA(1) process \(y_t=ε_t+αε_{t-1}\) :
\[Var\left( y_t \right)=(1+α^2)σ^2\]
- Result: The covariances of an MA(1) process \(y_t=ε_t+αε_{t-1}\) :
\[Cov\left( y_t,y_{t-1} \right)=ασ^2\]
- and
\[Cov\left( y_t,y_{t-k} \right)=0\]
- if \(k>1\) .
- Result: The correlations (or autocorrelations) of an MA(1) process \(y_t=ε_t+αε_{t-1}\) :
\[corr\left( y_t,y_{t-1} \right)= \frac{α}{1+α^2}\]
- \(corr\left( y_t,y_{t-k} \right)=0\) if \(k>1\) .