ABSTRACT
This chapter presents some ideas and approaches for some iterative model-fitting, using a univariate time series. Most parts of the approach are well known, in isolation; the aim is to combine them into a workable package. In fact, some iterative model-fitting is also needed in estimating the deterministic component, since the significance of a fitted term is only roughly known in advance of a complete analysis of the stochastic residual. Desirably, one needs some iterative process in which an initial transformation is estimated and then adjusted to suit the data better. Economic time series are not stationary, and do not satisfy the additive assumption that the time series is the sum of a deterministic component and a stochastic component. Most of the theory assumes this additivity. Since it usually does not hold, a transformation of the data is needed to ensure approximate additivity.
