Advanced ARIMA Modeling for Forecasting and Seasonal Adjustment
Purpose:
ARIMA models are mathematical models of the autocorrelation in a time series.
They are used for forecasting time series in many different fields. This course
is designed for users of seasonal adjustment and forecasting software who would like
a deeper understanding of ARIMA models and the Box-Jenkins method.
Because the focus is on forecasting for seasonal adjustment and ARIMA-model-based
adjustment as with SEATS, we will discuss only univariate time series.
The course is both practical and theoretical.
The course is best when taught with in-class computer work
(using either TRAMO or X-12-ARIMA), but could also be taught as
lectures only.
Duration: 3 days
Target Audience:
This course is intended for persons with a background in econometrics or
statistics who are interested in learning more about the details of
ARIMA modeling and the diagnostics involved. Since ARIMA modeling
is an important part of using SEATS, this course is especially useful
to TRAMO/SEATS users. The course is limited to 10 persons.
Prerequisites:
None, but topics require some knowledge of statistics, for example, the
participants should understand terms like mean, normal distribution,
covariance, and linear regression.
Topics Covered:
The course examines the following topics:
- ARMA processes
- Box-Jenkins method for ARIMA modeling
- Models for time series
- Model identification
- Fitting ARIMA models
- Estimation methods
- Regression models with ARIMA errors (regARIMA models)
- Detecting and removing outliers
- Detecting trading day and moving holiday effects
- Forecasting methods and evaluating forecasting performance
- Spectral methods
- Linear filters for seasonal adjustment
The course will involve the practical application of concepts
through the use of case studies, group discussion, and exercises.
Windows® is a registered trademark of Microsoft Corporation
Back to the Top
Last update: 10 January 2007
|
