Financial Time Series Tutorial 2024
Uploaded by
Sand RukshanFinancial Time Series Tutorial 2024
Uploaded by
Sand RukshanCommon questions
Powered by AIThe variance of the sum of two random variables X and Y is given by Var(X + Y) = Var(X) + Var(Y) + 2Cov(X,Y). Given Var(X) = 9 and Var(Y) = 4, plus Cov(X,Y) = Corr(X,Y) * sqrt(Var(X) * Var(Y)) = 0.25 * sqrt(9 * 4) = 0.25 * 6 = 1.5 . Therefore, Var(X + Y) = 9 + 4 + 2*1.5 = 16 .
The process Yt = 2Yt−1 − Yt−2 + et represents an ARIMA(2,0,0) model because it includes two autoregressive terms (2Yt−1 − Yt−2) and no differencing or moving average components, so p = 2, d = 0, q = 0. The parameters are ϕ1 = 2, ϕ2 = -1 .
For the series Yt = 5 + 2t + Xt, where Xt is a zero-mean stationary series, the mean function is E(Yt) = E(5 + 2t + Xt) = 5 + 2t, since E(Xt)=0 .
Cov(X, X + Y) = Cov(X, X) + Cov(X, Y) = Var(X) + Cov(X, Y). With Var(X) = 9 and Cov(X,Y) = 1.5, Cov(X, X + Y) = 9 + 1.5 = 10.5 .
For stationarity in an AR(p), the roots of the characteristic equation must lie outside the unit circle. For invertibility in an MA(q), the roots of the moving average polynomial must also lie outside the unit circle. For Yt = 2Yt−1 − Yt−2 + et, the AR part gives roots from 1-2z+z^2=0, which are 1, 1, indicating a unit root and lack of stationarity .
The given model Yt = Yt−1 − 0.25Yt−2 + et − 0.1et−1 is an ARIMA(2,0,1) model. Here, p = 2, d = 0, q = 1 with AR coefficients ϕ1 = 1, ϕ2 = -0.25 and MA coefficient θ1 = -0.1 .
Adding a deterministic trend to a zero-mean stationary component, like in Yt = 5 + 2t + Xt, doesn't change the component's autocovariance. The autocovariance of Xt remains unchanged as a zero-mean stationary series; however, the combined series Yt isn't stationary due to the trend, impacting long-term statistics .
To derive the autocorrelation function from first principles for Yt = 5 + et − 1/2et−1 + 1/4et−2, recognize Yt as MA(2), as it depends on current and past two noise terms. Assuming et is white noise with mean zero and variance σ², the autocovariance is γ(k) = 0 for k > 2, with the initial value γ(0) = σ²(1^2 + (1/2)^2 + (1/4)^2) and γ(1) = -0.5σ², γ(2) = 0.25σ². Normalize each by γ(0) to find the autocorrelations .
A time series is considered stationary if its statistical properties, such as mean, variance, and autocovariance, do not change over time. In the given time series Yt = 5 + 2t + Xt, Xt is a zero-mean stationary series, but the presence of the term 2t implies a deterministic trend, causing the mean function E(Yt) = E(5 + 2t + Xt) = 5 + 2t , which changes over time. Thus, Yt is not stationary due to the linear trend term 2t .
For the ARMA(1,1) process Xt = -0.6Xt-1 + Zt + 0.3Zt-1, where Zt is white noise, stationarity requires the AR coefficient's absolute value to be less than 1. Here, |−0.6| < 1, so it is stationary. Invertibility requires that the MA part, represented by the roots of (1 + 0.3B), lie outside the unit circle. Since 0.3 is within acceptable bounds for invertibility, the ARMA(1,1) process is both stationary and invertible .