Stats 349 – Stats 846 Assignment 1 -Solutions I Suppose that the stationary time series {xt : t T} satisfies the equation: xt = ut ut 1 ut 2 ... ut m m 1 where {ut : t T} is a white noise time series with variance 2 = 4.0. Show that the autocorrelation function of {xt : t T} is m 1 h m 1 x(h) = 0 0hm hm Solution: This time series is an MA(m) time series with 0 1 m 1 m 1 Thus h mh 2 i 0 i ih 1 m h 1 m 1 2 2 if 0 h m, h 0 if h m a nd 2 1 h m 1 m h 1 if 0 h m, h 0 if h m h 2 0 m 1 1 2 m 1 m 1 II Suppose that the stationary time series {xt : t T} satisfies the equation: xt = xt-3 + ut. where {ut : t T} is a white noise time series with variance 2 = 9.0. Find the autocorrelation function of {xt : t T}. Plot the autocorrelation function with = 0.8. Solution: This time series is an MA(3) time series with 1 2 0 and 3 . The Yule-Walker equations: 1 1 2 1 3 2 1 2 2 m h 1 2 11 2 3 1 or 2 1 3 1 2 2 1 3 3 Thus 1 2 1 2 1 and 1 2 1 0 and 1 0 unless 1 Also 2 1 0, Thus the solutions to the Yule Walker equations are 1 2 0 and 3 . For h > 3 we use h = 1h – 1+ 2h - 2+ 3h – 3 = h – 3 Thus 4 5 0, 6 2 , 7 8 0, 9 3 , etc. Page 1 Stats 349 – Stats 846 Assignment 1 -Solutions Plot of autocorrelation function: 1.2 1 0.8 0.6 0.4 0.2 40 38 36 34 32 30 28 26 24 22 20 18 16 14 12 10 8 6 4 2 0 0 III. Consider the Autoregressive (AR(3)) process of order 3 satisfying the equation: xt= 1xt-1 + 0.25xt-2 - 0.10xt-3+ + ut where var(ut) = 2. Suppose that the autocorrelation function, (h), at lag h = 1 takes on the values (1) = 0.50. In addition var(xt) = 10 and E[xt] = = 25 a) Determine the values of 1, and 2 . Solution This is an AR(3) time series with 1 (unknown), 2 = 0.25 and 3 = 0.10. The Yule-walker equations are 0.5 1 0.25 0.5 0.10 2 1 1 2 1 3 2 2 11 2 3 1 or 2 1 0.5 0.25 0.10 0.5 3 1 2 2 1 3 3 12 0.25 0.5 0.10 The first two equations can be used to solve for 1 and 2. 1 0.375 0.102 and 2 0.51 0.30 0.5 0.375 0.10 2 0.30 0.4875 0.46429 1.05 and 1 0.375 0.102 0.375 0.10 0.46429 0.32857 1.05 2 0.5 0.375 0.30 0.4875 and 2 Finally 25 1 2 3 0.32857 0.25 0.10 0.32143 and 0.32143 25 8.036 Now 1 0.5, 2 0.46429 . The third equation of the Yule-Walker equations can be used to calculate 3. Namely 3 12 2 1 3 0.32857 0.46429 0.25 0.50 0.10 0.377551 Page 2 Stats 349 – Stats 846 Assignment 1 -Solutions To compute 2 we use 10.0 0 2 1 1 1 2 2 3 3 2 1 0.32857 .5 0.25 0.46429 0.10 0.377551 2 0.681888 Thus 10.0 0.681888 6.81888 2 b) Compute and plot a graph of (h), the auotocovariance function of the process at lag h, for h = 0, 1, 2, 3, 4. Solution: For h >3 we use h 1h1 2 h2 3 h3 0.32857 h1 0.20h2 0.10h3 Thus 4 0.32857 0.377551 0.25 0.46429 0.10 0.5 0.290124 Finally The autocovariance function is determined by (h) = (0) h= 10.0 h Thus (0)=10.0, (1) = 5.0, (2) = 4.6429, (3) = 3.77551, (4) = 2.90124 IV. An AR(2) time series, {xt : t T}, if | r1 | >1 and | r2 | >1, where r1 and r2 are the roots ofthe polynomial 1 -1 x - 2x2 = 0. Show that an AR(2) time series is stationary if the parameters, 1 and 2, satisfy the following inequalities: 2 + 1 < 1, 2 - 1 < 1, -1 < 2 < 1. 2 Show that roots r1 and r2 are complex if 1+ 4 2 < 0. Graph this regions implied by these inequalities. Solution: x 1 1 1 2 x 1 1 x 2 x 2 1 1 1 x x where r1 r2 r1 r2 r1r2 1 4 2 r1 , r2 are the roots of 1 1 x 2 x 2 2 2 Note: r1 , r2 are complex if 4 2 0 Also 1 1 1 1 1 and 2 , , 2 r1 r2 r1r2 r1 r2 1 1 1 1 1 1 1 r1 r2 r1r2 r1 r2 1 1 1 1 1 and 1 2 1 1 1 1 r1 r2 r1r2 r1 r2 1 Now 2 1 , 1 1 2 0 and 1 2 1 0 if and only if r1 1 and r2 1 r1 r2 Thus 1 1 2 1 Page 3