ABSTRACT
There are two types of nonstationary time series, one with a drifting mean value and the other with a varying structure around the mean value function. For the latter type of nonstationary time series, the variance, the autocovariance function, or the power spectrum of the time series change over time. This chapter presents two methods for the analysis of such nonstationary time series. One is an estimation method for time-varying variance, and the other is a method for modeling the time-varying coefficient aggressive model. The estimation of the stochastic volatility in financial time series analysis can be considered as a typical example of the estimation of the time-varying variance. The characteristics of stationary time series can be expressed by an autocovariance function or a power spectrum. Therefore, for nonstationary time series with a time-varying stochastic structure, it is natural to consider that its autocovariance function and power spectrum change over time.
