ABSTRACT

In this chapter we introduce the unobserved components model and describe the genesis of its main components.

3.1 The unobserved components model

A natural way humans tend to think about time series is as the sum of non directly observable components such as trend, seasonality, cycle. The seminal works of Holt (1957) and Winters (1960) on exponentially weighted moving average (EWMA) methods are early successful examples of this approach. Also official statisticians are used to conceive time series as sum of unobserved components: since the appearance of the first intra-annual time series, national statistical institutes had to provide seasonally adjusted versions of them.