tailieunhanh - Econometric theory and methods, Russell Davidson - Chapter 13

Chapter 13 Methods for Stationary Time-Series Data Introduction Time-series data have special features that often require the use of specialized econometric techniques. We have already dealt with some of these. | Chapter 13 Methods for Stationary Time-Series Data Introduction Time-series data have special features that often require the use of specialized econometric techniques. We have already dealt with some of these. For example we discussed methods for dealing with serial correlation in Sections through and in Section and we discussed heteroskedasticity and autocorrelation consistent HAC covariance matrices in Section . In this chapter and the next we discuss a variety of techniques that are commonly used to model and test hypotheses about economic time series. A first point concerns notation. In the time series literature it is usual to refer to a variable series or process by its typical element. For instance one may speak of a variable yt or a set of variables Yt rather than defining a vector y or a matrix Y. We will make free use of this convention in our discussion of time series. The methods we will discuss fall naturally into two groups. Some of them are intended for use with stationary time series and others are intended for use with nonstationary time series. We defined stationarity in Section . Recall that a random process for a time series yt is said to be covariance stationary if the unconditional expectation and variance of yt and the unconditional covariance between yt and yt-j for any lag j are the same for all t. In this chapter we restrict our attention to time series that are covariance stationary. Nonstationary time series and techniques for dealing with them will be discussed in Chapter 14. Section discusses stochastic processes that can be used to model the way in which the conditional mean of a single time series evolves over time. These are based on the autoregressive and moving average processes that were introduced in Section . Section discusses methods for estimating this sort of univariate time-series model. Section then discusses singleequation dynamic regression models which provide richer ways to .