This tutorial will introduce the attendees to the analysis and forecasting of time series by state space methods using R. State space models provide a very flexible framework that has proved highly successful in analysing data arising in a wide array of disciplines, such as, to mention a few, economics, business and finance, engineering, physics, hydrology, and climatology. Part of their appeal resides in the construction of an easily interpretable, smoothly evolving process underlying the noisy observations, and in the ability of the model to smooth and forecast this process in addition to the data. The goal of the tutorial is to give the attendee the conceptual tools to set up a meaningful state space model for a specific time series together with the computational skills needed to estimate the unknown parameters of the model, smooth and forecast the series.
Several contributed packages are currently available to estimate and analyse state space models. Unfortunately, different packages use different user interfaces and even different conventions about the model definition; moreover, at least for some of the packages, the user interface is fairly austere and the help system somehow underdeveloped, two aspects that may be intimidating especially to the novice. This tutorial will provide clear guidance on the most important functions in the different packages, highlighting strengths and weaknesses of each of them.