PSTAT 174/274: Assignment # 2
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
PSTAT 174/274: Assignment # 2. (Released 27th Feb - Due 5th March 11:55pm PST)
• All Questions are equally weighted and this worksheet contributes 15% to overall grade. (if included in best 2 out of 3)
• To get full marks you must show all working and justify your solution/discussion or interpret your solution where appropriate.
• If the question is a code question you must comment all code appropriately and present all outputs with correct labelling of plots and discussion/captions.
1. (174 & 274 attempt) Characteristic Polynomials and Stationarity. (Total marks for question 33%)
For each of the time series models below undertake the following steps:
• Write each time series in characteristic polynomial form using backshift operator (MA or AR or ARMA polynomials as required)
• Factorise the relevant polynomials and find the roots of the polynomials.
• You are given the following additional information regarding the relationship between stationarity and the characteristic polynomial roots.
– An AR(p) process is a stationary process if and only if the modulus of all the roots of the AR characteristic polynomial are greater than one.
– An MA(q) process is always stationary if the MA coefficients are finite.
– An ARMA(p,q) model admits a unique stationary solution when the AR characteristic poly- nomial and the MA characteristic polynomial have no common roots and the AR polynomial satisfies that it has no unit roots.
• Use this information to determine which of the below models is stationary.
(Make sure to justify your answer.)
(a) ∇Yt + (Yt + 5Yt − 1 ) = ϵt − ϵt − 1 − ϵt −2 + (1 + B)∇Yt − Yt
(b) Yt − − 2Yt − 1 = 2ϵt + ϵt − 1 − (1 − B2)ϵt
(c) Yt+1 + Yt − (1 − B)(1 + B)Yt+1 = ∇ϵt
(d) ϕ(B )Yt = ∇2 ϵt + (1 + 2B)ϵt ,
where one of the roots of the polynomial ϕ(B ) lies inside the unit circle and all other roots lie outside the unit circle.
(e) Let {Yt} be a stationary process with ACF ρ(·) and let
M := E((Yt − (ϕk1Yt − 1 + . . . + ϕkkYt −k ))2 ) .
By differentiating M and setting the result to zero, show that the values ϕk1 , . . . , ϕkk which minimise
M satisfy the Yule-Walker equations
ρ(j) = ϕk1ρ(j − 1) + . . . + ϕkkρ(j − k)
for j = 1, . . . , k .
2. (174 & 274 attempt) ARMA and ARIMA Models.
(Total marks for question 33%)
(a) (ONLY 174 Attempt) Let {Yt} be a stationary ARMA(1, 1) model with mean µ 0 and white noise sequence {ϵt} ∼ WN(0, σ2 ). Calculate E (ϵtYt+k), for all k ≥ 0 and express your answer in
terms of Kronecker-Delta functions where possible.
(b) (ONLY 274 Attempt)
Given Y0 = c where c is a real positive constant, find the variance Var (Yt) and covariance Cov (Yt, Yt+k) of a process {Yt} given by an ARIMA(0,1,0) model which can be written as
∇Yt = ϵt .
Comment on whether Yt is weakly stationary.
3. (174 & 274 attempt) Time Series Regression Models.
(Total marks for question 33%)
(a) Consider a time series regression model given as follows
Yt = β0 + β1Xt + t
where
t = ϕ 1 t − 1 + ϕ2 t −2 + ϵt ,
where ϵt ∼ WN (0, σ2 ).
Furthermore, you are told that ϕ 1 = 0.1, ϕ2 = −0.1 and σ 2 = 1.
Derive the least squares estimator for the coefficients β0 and β1 . Your solution should obtain estimators ex- pressed in terms of the data Y = (Y1 , . . . , YT )T and the parameters ϕ1 , ϕ2 and covariates X = (X1 , . . . , Xt)T and should account appropriately for the non i.i.d. regression errors t , such that the resulting estimators you obtain are unbiased.
(b) Consider the time series regression model
Yt = α + β0Xt + β1Xt − 1 + β2Xt −2 + ϵt
Consider two scenarios:
• Scenario I: There is an instantaneous shock in {Xt} that only occurs at time t = s (is not a permanent change in Xt)
• Scenario II: There is an permanent shock in {Xt} that first occurs at time t = s and creates a permanent change in Xt that continues for all times t ≥ s.
For each scenario derive the short run, intermediate and long run effect of the shock in Xt on the regression response Yt in terms of the regression coefficients α, β0 , β 1 and β2 .
2023-03-04