ST414 – Spectral Analysis of Time Series Data Lecture 1 28 January 2014 Examples 2 Brockwell and Davis, 2002 Examples 3 Shumway & Stoffer, 2004 Examples 4 Ombao et al, 2001 Examples 5 Kakizawa et al, 1998 Examples 6 Ombao et al, 2001 Examples 7 Examples 8 Shumway & Stoffer, 2004 Objective Develop a working knowledge of statistical theories and methodologies for spectral analysis of time series data. 9 Today’s Objectives • Discuss periodicity • Discuss basic time series models 10 Periodicity 11 Shumway & Stoffer, 2004 Periodicity where 12 Periodicity More generally, where What is Var(X(t))? 13 Periodicity 14 Shumway & Stoffer, 2004 Preliminaries X(t) is said to be strictly stationary if the probabilistic behaviour of every collection of values is identical to that of the shifted set 15 Preliminaries X(t) is said to be weakly stationary if 1) E(X(t)) is invariant with respect to t and 2) πΆππ£ π π‘ , π π = πΎ(β), where h = |s-t|. πΎ(β) is called the autocovariance function of X(t). 16 Preliminaries 1. πΎ 0 ≥ 0 2. |πΎ β | ≤ πΎ 0 3. πΎ β =πΎ −β 17 Preliminaries The autocorrelation function of a weakly stationary time series X(t) is πΎ(β) π β = πΎ(0) 18 White Noise X(t) is white noise if it is a collection of uncorrelated random variables identically distributed with mean 0 and finite variance π 2. Is X(t) weakly stationary? What is πΎ(β)? What is π β ? 19 White Noise 20 White Noise 21 MA(1) π π‘ = π π‘ + ππ(π‘ − 1) π π‘ is white noise Is X(t) weakly stationary? What is πΎ(β)? What is π β ? 22 MA(1) 23 MA(1) 24 AR(1) π π‘ = ππ π‘ − 1 + π π‘ π π‘ is white noise Is X(t) weakly stationary? What is πΎ(β)? What is π β ? 25 AR(1) 26 AR(1) 27 AR(1) 28 More Preliminaries X(t) is a linear process if it has the representation ∞ π π‘ = ππ π(π‘ − π), π=−∞ for all t, where Z(t) is white noise and {ππ } is a sequence of constants with ∞ π=−∞ |ππ | < ∞. 29 More Preliminaries If X(t) ~ AR(1), then X(t) has a MA ∞ representation: ∞ π π π(π‘ − π), π π‘ = π=0 Is X(t) weakly stationary? What is πΎ(β)? What is π β ? 30 MA(2) Consider the MA(2) model: π π‘ = π π‘ + π1 π π‘ − 1 + π2 π(π‘ − 2) π π‘ is white noise 2−|β| πΎ β = π2 ππ ππ+ β , if |β| ≤ 2 π=0 0, where π0 = 1. if β > 2 31 ACF of MA(2) 32 ACF of AR(1) 33 The PACF Heuristically, take the correlation between X(t) and X(s) with the linear effect of everything in between removed. 34 ACF of AR(1) 35 PACF of AR(1) 36 PACF of AR(2) 37 Example 38 Example 39 Example 40 An Exercise Let π π‘ = π1 cos 2πππ‘ + π2 sin 2πππ‘ where π1 , π2 iid 0, π 2 . Is X(t) weakly stationary? If so, what is πΎ(β)? 41