ARMA ARIMA Univariate Time Series Models • Where we attempt to predict returns using only information contained in their past values. Some Notation and Concepts • A Strictly Stationary Process A strictly stationary process is one where P{y ≤ b ,..., y ≤ b } = P{y ≤ b ,..., y ≤ b } t 1 t n t +m 1 t +m n 1 n 1 n i.e. the probability measure for the sequence {y } is the same as that for {y } ∀ m. t t+m • A Weakly Stationary Process If a series satisfies the next three equations, it is said to be weakly or covariance stationary 1. E(y ) = µ, t = 1,2,...,∞ t 2 2. E(y − µ)(y − µ) =σ < ∞ t t 3. E(y −µ)(y −µ) =γ ∀ t , t t t t −t 1 2 1 2 2 1 Univariate Time Series Models (cont’d) • So if the process is covariance stationary, all the variances are the same and all the covariances depend on the difference between t and t . The moments 1 2 E(y − E(y ))(y − E(y )) =γ , s = 0,1,2, ... t t t+s t+s s are known as the covariance function. • The covariances, γ, are known as autocovariances. s • However, the value of the autocovariances depend on the units of measurement of y . t • It is thus more convenient to use the autocorrelations which are the autocovariances normalised by dividing by the variance: γ , s = 0,1,2, ... τ = s s γ 0 • If we plot τ against s=0,1,2,... then we obtain the autocorrelation function or s correlogram. A White Noise Process • A white noise process is one with (virtually) no discernible structure. A definition of a white noise process is E(y ) = µ t Var(y ) =σ2 t ⎧σ2 if t = r γ = ⎨ t−r ⎩0 otherwise • Thus the autocorrelation function will be zero apart from a single peak of 1 at s = 0. τ ∼ approximately N(0,1/T) where T = sample size s • We can use this to do significance tests for the autocorrelation coefficients by constructing a confidence interval. 1 ±.196× • For example, a 95% confidence interval would be given by . If T the sample autocorrelation coefficient, τ (cid:3) , falls outside this region for any s value of s, then we reject the null hypothesis that the true value of the coefficient at lag s is zero. Joint Hypothesis Tests • We can also test the joint hypothesis that all m of the τ correlation coefficients k are simultaneously equal to zero using the Q-statistic developed by Box and m Pierce: Q = T∑τ2 k k=1 where T = sample size, m = maximum lag length • The Q-statistic is asymptotically distributed as a χ 2 . m • However, the Box Pierce test has poor small sample properties, so a variant has been developed, called the Ljung-Box statistic: m τ2 Q∗ = T(T + 2)∑ k ~ χ2 m T − k k=1 • This statistic is very useful as a portmanteau (general) test of linear dependence in time series. An ACF Example • Question: Suppose that a researcher had estimated the first 5 autocorrelation coefficients using a series of length 100 observations, and found them to be (from 1 to 5): 0.207, -0.013, 0.086, 0.005, -0.022. Test each of the individual coefficient for significance, and use both the Box- Pierce and Ljung-Box tests to establish whether they are jointly significant. • Solution: A coefficient would be significant if it lies outside (-0.196,+0.196) at the 5% level, so only the first autocorrelation coefficient is significant. Q=5.09 and Q*=5.26 Compared with a tabulated χ2(5)=11.1 at the 5% level, so the 5 coefficients are jointly insignificant. Moving Average Processes • Let u (t=1,2,3,...) be a sequence of independently and identically t distributed (iid) random variables with E(u )=0 and Var(u )=σ 2 , then ε t t y = µ+ u + θu + θu + ... + θu t t 1 t-1 2 t-2 q t-q is a qth order moving average model MA(q). • Its properties are E(y )=µ; Var(y ) = γ = (1+θ 2 + θ 2 + . . . + θ 2 )σ2 t t 0 1 2 q Covariances ⎧(θ +θ θ +θ θ + ... +θθ )σ2 for s = 1,2,...,q ⎪ s s+1 1 s+2 2 q q−s γ = ⎨ s ⎪0 for s > q ⎩ Example of an MA Problem 1. Consider the following MA(2) process: X = u +θu +θ u t t 1 t−1 2 t−2 where ε is a zero mean white noise process with variance σ 2. t (i) Calculate the mean and variance of X t (ii) Derive the autocorrelation function for this process (i.e. express the autocorrelations, τ, τ, ... as functions of the parameters θ and θ). 1 2 1 2 (iii) If θ = -0.5 and θ = 0.25, sketch the acf of X . 1 2 t Solution (i) If E(u )=0, then E(u )=0 ∀ i. t t-i So E(X ) = E(u + θu + θu )= E(u )+ θE(u )+ θE(u )=0 t t 1 t-1 2 t-2 t 1 t-1 2 t-2 Var(X ) = E[X -E(X )][X -E(X )] t t t t t but E(X ) = 0, so t Var(X ) = E[(X )(X )] t t t = E[(u + θu + θu )(u + θu + θu )] t 1 t-1 2 t-2 t 1 t-1 2 t-2 = E[ u 2 + θ 2 u 2 + θ 2 u 2 +cross-products] t 1 t−1 2 t−2 But E[cross-products]=0 since Cov(u ,u )=0 for s≠0. t t-s Solution (cont’d) So Var(X ) = γ = E [ u 2 + θ 2 u 2 + θ 2 u 2 ] t 0 t 1 t−1 2 t−2 = σ2 +θ2σ2 +θ2σ2 1 2 = (1+θ2 +θ2)σ2 1 2 (ii) The acf of X . t γ = E[X -E(X )][X -E(X )] 1 t t t-1 t-1 = E[X ][X ] t t-1 = E[(u +θu + θu )(u + θu + θu )] t 1 t-1 2 t-2 t-1 1 t-2 2 t-3 = E[( θ u 2 + θ θ u 2 )] 1 t−1 1 2 t−2 = θ σ2 +θθ σ2 1 1 2 = (θ +θθ )σ2 1 1 2
Description: