Few questions (White noise, PSD, MA, AR, Yule Walker, PACF,Levinson-
Durbin)
1. Define white noise and list its main properties. How does it differ from colored noise?
2. If a process e[k] is white noise with mean zero and variance σ2, calculate the
autocorrelation function (ACF) for this process.
3. Explain the role of white noise in generating other time series processes, such as AR and
MA models.
4. What is the power spectral density (PSD), and how does it relate to the autocorrelation
function?
5. Given a white noise process with variance σ2, determine the PSD.
7. Define a Moving Average (MA) model of order q and write its general form.
8. For the MA(1) process y[k]=e[k]+θe[k−1], where e[k] is white noise with variance σ2,
derive the ACF.
9. Explain the conditions under which an MA process is invertible.
10.Define an Autoregressive (AR) model of order p and write its general form.
11.For the AR(1) process y[k]=ϕy[k−1]+e[k], where e[k] is white noise with variance σ2,
derive the ACF.
12.Explain the concept of stationarity in the context of an AR(1) process. What conditions must
ϕ satisfy for the process to be stationary?
13.What are the Yule-Walker equations, and how are they derived?
14.For an AR(2) process y[k]=ϕ1y[k−1]+ϕ2y[k−2]+e[k], write the Yule-Walker equations.
15.Explain how the Yule-Walker equations are used to estimate the parameters of an AR
process.
16.Define the partial autocorrelation function (PACF) and explain how it differs from the
ACF.
17.Describe how the PACF behaves for an AR(p) process. How can the PACF help in
identifying the order p of an AR model?
18.Given the PACF for a time series, how would you determine if the process is AR(1), AR(2),
or MA?
Few Questions to Think....
1. Consider a MA(2) process defined by:
y[k]=e[k]+0.5e[k−1]−0.3e[k−2]
where e[k] is white noise with zero mean and variance σ2=1.
1. Find the theoretical autocorrelation function (ACF) of y[k] for lags m=0,1,2,3.
2. Sketch the ACF for this MA(2) process.
2. Given an MA(1) process:
y[k]=e[k]+θe[k−1]
where e[k] is white noise with variance σ2=2.
1. Derive an expression for the ACF of y[k] for lags m=0 and m=1.
2. If the observed ACF at lag 1 is 0.5, estimate the parameter θ of this MA(1) process.
3. Consider an AR(1) process defined by:
y[k]=0.7y[k−1]+e[k]
where e[k] is white noise with variance σ2=1.
1. Derive the theoretical ACF for y[k] for lags m=0,1,2,3.
2. Explain how the value of the AR coefficient (0.7) affects the behavior of the ACF as m
increases.
4. For an AR(2) process:
y[k]=1.5y[k−1]−0.75y[k−2]+e[k]
where e[k] is white noise with variance σ2=1.
1. Using the Yule-Walker equations, set up the system of equations for the autocorrelation
function Ry(m) at lags m=0,1,2.
2. Solve these equations to find Ry(0), Ry(1), and Ry(2).
5. Consider an AR(1) process given by:
y[k]=0.6y[k−1]+e[k]
where e[k] is white noise with variance σ2=1.
1. Derive the Yule-Walker equation for this AR(1) process.
2. Using the Yule-Walker equation, calculate the theoretical value of Ry(0) and Ry(1).
6. For an AR(2) process:
y[k]=0.8y[k−1]+0.3y[k−2]+e[k]
where e[k] is white noise with variance σ2=1.
1. Write down the Yule-Walker equations for this AR(2) process.
2. Solve these equations to find Ry(0), Ry(1), and Ry(2).
7. Suppose you are given the autocorrelation values for a stationary process: R(0)=4, R(1)=2,
R(2)=1.
1. Use the Levinson-Durbin algorithm to compute the first-order and second-order
autoregressive coefficients ϕ1 and ϕ2.
2. Calculate the prediction error variance for this process up to order 2.
8. For an unknown process, the following autocorrelation values are provided: R(0)=10,
R(1)=6, R(2)=2, R(3)=1.
1. Apply the Levinson-Durbin algorithm to determine the AR(3) coefficients.
2. Compute the prediction error variance up to order 3 using the results from the Levinson-
Durbin recursion.
