ABSTRACT
The specific to general (SG) approach to time series analysis implicitly assumes that the means and the variances of the economic variables in a regression model remain constant over time. That is, they are time-invariant, and when these assumptions break down, alternative methods of estimation to the OLS are used to generate efficient estimators. Within this framework the underlying reasons for the breakdown of these assumptions are seldom systematically investigated. The mean and variances of many economic and financial variables are, however, time-variant and their time series are non-stationary. The non-stationary time series and their treatment have given rise to some important contributions in econometrics in recent years, including unit root tests and cointegration analysis. In the next three chapters we will focus attention on these topics. Each chapter will include an extended summary of key issues to help with better understanding of techniques and applications.
