© 2020 KAB Tutorial Sheet 1.0 1. A stationary π΄π (2) process is defined by the equation: 5 1 ππ‘ = ππ‘−1 − ππ‘−2 + ππ‘ 6 6 Determine the values of ππ and ππ for π = 1, 2, 3, …. 2. Determine the characteristic polynomial of the process defined by the equation: ππ‘ = 5 − 2(ππ‘−1 − 5) + 3(ππ‘−2 − 5) + ππ‘ and calculate its roots. Hence comment on the stationarity of the process. 3. Determine whether the process ππ‘ = 2 + ππ‘ − 5ππ−1 + 6ππ−2 is invertible. 4. Given that λ = 2 is a root of the characteristic equation of the process: 11 1 ππ−1 − ππ−2 + ππ−3 + ππ 6 3 calculate the other roots and classify the process as πΌ(π). ππ = 5. Calculate ππ , π = 0,1,2,3, . .. for the process: ππ = ππ − ππ−1 + 0.25ππ−2 + 3 where ππ is a white noise process with mean 0 and variance 1. 6. {ππ‘ } is a white noise process with mean 0 and variance π 2 . Calculate the following: (a) πππ£(π2 , π3 ) (b) πππ£(π3 , π3 ) (c) πππ£(π1 + π2 , π1 + π3 ) 1 (d) πππ£(0.5π1 + π2 , 0.5π2 + π3 ) (e) πππ(π1 + 4 π2 ) 7. Let ππ‘ be a sequence of independent standard normal random variables. Determine which of the following processes are stationary ti me series. (i) ππ‘ = sin(ππ‘ + π) , where π is uniformly distributed on the interval [0,2π] (ii) ππ‘ = sin(ππ‘ + ππ‘ ) (iii) ππ‘ = ππ‘−1 + ππ‘ (iv) ππ‘ = ππ‘−1 + ππ‘ (v) ππ‘ = 2 + 3π‘ + 0.5ππ‘−1 + ππ‘ + 0.3ππ‘−1 8. Give an expression for 2ππ‘ − 5ππ‘−1 + 4ππ‘−2 − ππ‘−3 in terms of second order differences. 9. A time series is defined by the relationship ππ‘ = ππ‘−1 + ππ‘ , where the ππ‘ are πΌπΌπ· π(0, π 2 ) random variables. Determine the relationship between π£ππ(ππ‘ ) and π£ππ(ππ‘−1 ), and hence comment on the stationarity of this series. 1 10. Determine whether the process ππ = ππ−1 − 2 ππ−2 is stationary. 11. Determine whether the following time series is stationary and/or invertible: Page 1 of 4 © 2020 KAB ππ‘ − 0.1ππ‘−1 + 0.2ππ‘−2 = ππ‘ + 0.2ππ‘−1 where {ππ‘ } represents a set of uncorrelated random variables with mean 0 and variance π 2. 12. An autoregressive stationary time series ππ‘ is defined by the relationship: ππ‘ = 0.6ππ‘−1 + 0.4ππ‘−2 − 0.1ππ‘−3 + ππ‘ for integer times π‘, where {ππ‘ } represents a set of uncorrelated random variables with mean 0 and variance π 2 . (i) Explain why πππ£(ππ‘ , ππ‘+1 ) = 0 and πππ£(ππ‘−1 , ππ‘ ) = πππ£(ππ‘−1 , ππ‘−2 ). (ii) By considering πππ£(ππ‘ , ππ‘−π ).when π = 0,1,2,3, write down a set of four equations relating the values of the autocovariance function πΎπ at lags π = 0,1,2,3,. (iii) Solve the four equations in part (ii) to find both the autocovariance function and the autocorrelation function for lags 0, 1, 2 and 3. 13. Calculate the autocorrelation function of the process ππ = 1 + ππ − 5ππ−1 + 6ππ−2 . 14. (i) The first differences of a time series π can be modelled by the process: ∇ππ = 0.5∇ππ−1 + ππ Determine the model for ππ . (ii) Show that the process ππ is non-stationary. 15. (i) Show that the relationship ππ‘ = 0.7ππ‘−1 + 0.3ππ‘−2 + ππ‘ + 0.7ππ‘−1 . (where the π’π denote white noise) defines an π΄π πΌππ΄(1,1,1) process. (ii) Show carefully that the relationship ππ‘ = 1.5ππ‘−1 + 0.3ππ‘−3 + ππ‘ + 0.5ππ‘−1 cannot be expressed as an π΄π πΌππ΄(1,2,1) process. 16. Consider the process with defining equation: ππ = 5ππ−1 − 0.4ππ−2 + ππ−3 + ππ Write this as a vector process that possesses the Markov property. 17. Let ππ = ππ + ππ−2 be an ππ΄(2) process where ππ ~π(0,1). (i) Calculate π(ππ ≥ 0|ππ−1 ≤ 0) (ii) Compare your answer to (i) with π(ππ ≥ 0|ππ−1 ≤ 0, ππ−2 ≤ 0)and hence comment on whether the process is Markov. 18. Calculate the values of π1 and π2 , the autocorrelation function at lags 1 and 2, for the stationary π΄π (2) process defined by the equation: ππ = −0.8ππ−1 + 0.1ππ−2 + ππ Page 2 of 4 © 2020 KAB 19. Consider the time series model defined by: ππ‘ = πΌ1 ππ‘−1 + πΌ2 ππ‘−2 + πΌ3 ππ‘−3 + ππ‘ where ππ‘ is white noise. (i) Show that the autocorrelation coefficient with lag 1 for this process is: πΌ1 + πΌ2 πΌ3 π1 = 1 − πΌ2 − πΌ1 πΌ3 − πΌ32 (ii) Consider the case where πΌ1 = πΌ2 = πΌ3 = 0.2. (a) Comment on the stationarity of this model. π»πππ‘: 5 − π₯ − π₯ 2 − π₯ 3 ≈ (1.278 − π₯)(3.912 + 2.278π₯ + π₯ 2 ) (b) Calculate π1 and π2 , (c) Calculate the partial autocorrelation coefficients π1 and π2 , (d) Sketch correlograms of the autocorrelation function and the partial autocorrelation function. (You are not required to calculate the coefficients for higher lags.) 20. {ππ‘ } is a stationary π΄π ππ΄(1,2) time series defined at integer times by the relationship: ππ‘ = πΌππ‘−1 + ππ‘ + π½ππ‘−2 where πΌ, π½ are constants and {ππ‘ } is a purely random process with mean 0 and constant variance π 2 . (ii) Define the term ‘weakly stationary process’. (iii) Assuming that the above process has a very long history, state the conditions on πΌ and π½ needed to ensure that it is: (a) Stationary (b) Invertible. (iv) Show that for any integer π : πππ£(ππ , ππ ) = π 2 (v) πππ£(ππ , ππ −1 ) = πΌπ 2 πππ£(ππ , ππ −2 ) = (πΌ 2 + π½)π 2 (a) By considering πππ£(ππ‘ , ππ‘ ), πππ£(ππ‘ , ππ‘−1 ) and πππ£(ππ‘ , ππ‘−2 ), write down three equations involving πΎ0 , πΎ1 and πΎ2 . (b) Hence find expressions for πΎ0 , πΎ1 and πΎ2 in terms of the parameters πΌ, π½ and π 2. Page 3 of 4 © 2020 KAB (vi) (a) Calculate the values of π0 , π1 , π2 and π3 in the case where πΌ = −0.4 and π½ = −0.9. (b) Hence sketch a graph of the autocorrelation function ππ for lags π = 0, 1, 2, … , 10 in this case. …… Page 4 of 4