TimeSeriesBook.pdf

# R t t 1 100 102 200 c t 101 4 see also exercise 172

• 435

This preview shows pages 363–366. Sign up to view the full content.

R t = 0 , t = 1 , . . . , 100 , 102 , . . . , 200; c > 0 , t = 101. 4 See also exercise 17.2

This preview has intentionally blurred sections. Sign up to view the full version.

17.1. THE STATE SPACE REPRESENTATION 345 This means that W t = 0 and that Y t = X t for all t except for t = 101. For the missing observation, we have G 101 = Y 101 = 0. The variance for this observation is set to R 101 = c > 0. The same idea can be used to obtain quarterly data when only yearly data are available. This problem typically arises in statistical offices which have to produce, for example, quarterly GDP data from yearly observations incor- porating quarterly information from indicator variables (see Section 17.4.1). More detailed analysis for the case of missing data can be found in Harvey and Pierce (1984) and Brockwell and Davis (1991, Chapter 12.3). Structural Time Series Analysis An important application of the state space representation in economics is the decomposition of a given time series into several components: trend, cycle, season and irregular component. This type of analysis is usually coined structural time series analysis (See Harvey, 1989; Mills, 2003). Consider, for example, the additive decomposition of a time series { Y t } into a trend T t , a seasonal component S t , a cyclical component { C t } , and an irregular or cyclical component W t : Y t = T t + S t + C t + W t . The above equation relates the observed time series to its unobserved com- ponents and is called the basic structural model (BSM) (Harvey, 1989). The state space representation is derived in several steps. Consider first the case with no seasonal and no cyclical component. The trend is typically viewed as a random walk with time-varying drift δ t - 1 : T t = δ t - 1 + T t - 1 + ε t , ε t WN(0 , σ 2 ε ) δ t = δ t - 1 + ξ t , ξ t WN(0 , σ 2 ξ ) . The second equation models the drift as a random walk. The two distur- bances { ε t } and { ξ t } are assumed to be uncorrelated with each other and with { W t } . Defining the state vector X ( T ) t as X ( T ) t = ( T t , δ t ) 0 , the state and the observation equations become: X ( T ) t +1 = T t +1 δ t +1 = 1 1 0 1 T t δ t + ε t +1 ξ t +1 = F ( T ) X ( T ) t + V ( T ) t +1 Y t = (1 , 0) X ( T ) t + W t with W t WN(0 , σ 2 W ). This representation is called the local linear trend (LLT) model and implies that { Y t } follows an ARIMA(0,2,2) process (see exercise 17.5.1).
346 CHAPTER 17. KALMAN FILTER In the special case of a constant drift equal to δ , σ 2 ξ = 0 and we have that ∆ Y t = δ + ε t + W t - W t - 1 . { Y t } therefore follows a MA(1) process with ρ (1) = - σ 2 W / ( σ 2 ε + 2 σ 2 W ) = - (2 + κ ) - 1 where κ = σ 2 ε 2 W is called the signal-to-noise ratio. Note that the first order autocorrelation is necessarily negative. Thus, this model is not suited for time series with positive first order autocorrelation in its first differences. The seasonal component is characterized by S t = S t - d and d t =1 S t = 0 where d denotes the frequency of the data. 5 Given starting values S 1 , S 0 , S - 1 , . . . , S - d +3 , the following values can be computed recursively as: S t +1 = - S t - . . . - S t - d +2 + η t +1 , t = 1 , 2 , . . .

This preview has intentionally blurred sections. Sign up to view the full version.

This is the end of the preview. Sign up to access the rest of the document.
• Spring '17
• Raffaelle Giacomini

{[ snackBarMessage ]}

### What students are saying

• As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

Kiran Temple University Fox School of Business ‘17, Course Hero Intern

• I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

Dana University of Pennsylvania ‘17, Course Hero Intern

• The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

Jill Tulane University ‘16, Course Hero Intern