statrefs home‎ > ‎Main‎ > ‎Methods‎ > ‎

Time Series

Why time series modeling and forecasting?

Standard regression methods, like all statistical methods, have certain underlying assumptions that are required to be met for the model, and its results, are valid.  One of these assumptions is that the observations (rather, the residuals) are independent or uncorrelated.  When the observations cannot be said to be independent from each other, they are said to be autocorrelated.  Under these conditions, standard regression methods become less effective (the assumption of iid is not met, affecting the estimate of the error variance), and time series methods are needed.
Useful links to online references with Time Series methods include;

A reasonably good reference with an overview of select time series methods, but that doesn't go too deep into theoretical detail, is A Guide to Modern Econometrics (Verbeek).
Another good reference that goes a little deeper into the theory behind a few basic elements of time series applications is Statistical Control by Monitoring and Adjustment (Box).

Other references and papers are attached at the bottom of this page, or to sub-pages in this section.
Related fields include:
    • Time series for financial data has additional issues that must be considered.  One of the main issues is the highly volatile nature of most financial data that violates the assumptions of most standard regression methods and many of the more basic time series methods, creating questions about the validity of resulting models.
More detailed references, complete with the underlying theory, include:

Organizations with useful information about Time Series, Forecasting, and Econometrics:

  • Swiss Federal Institute of Technology (ETH) Zurich
    • Rmetrics - "The premier open source software solution for teaching and training quantitative finance" at

Other resources