Peering into the Future: Introduction to Time Series Methods for Forecasting
Presented: Wednesday September 5, 2018, 8:00am-11:30am
Presented by:
David A. Dickey is the William Neal Reynolds Distinguished Professor of Statistics at NC State University. He is a Fellow of the American Statistical Association and co-inventor of the Dickey-Fuller test that appears in most econometrics and time series textbooks and most time series computer packages including SAS PROC ARIMA. Dave received his PhD in 1976 from Iowa State University, and his research has been in time series analysis. He has been a contract instructor for SAS since 1981 and is a SAS Press author. His book, SAS for Forecasting Time Series, Third Edition, was published in March 2018. Dave is a frequent speaker at SAS Global Forum, and local and regional SAS user groups. At NC State he is a member of the Academy of Outstanding Teachers and the Academy of Outstanding Faculty Engaged in Extension. In addition to his home department of Statistics, Dave held associate appointments in Ag and Resource Economics, Financial Math, and the Integrated Manufacturing and Systems Engineering Institute. He was a founding faculty member in the NCSU Institute for Advanced Analytics and remains an affiliate of the institute.
Description:
SAS has a whole suite of tools for forecasting and analysis of data taken over time. While these procedures can have a different look and feel from procedure to procedure, they are for the most part based on autoregressive integrated moving average models or ARIMA models. Starting from the beginning, this tutorial explains what is meant by an ARMA model and shows how to identify and estimate an appropriate ARIMA model for your data complete with examples. Incorporation of polynomial trends and other types of inputs is then discussed
An important aspect of such analysis is the decision as to whether the original series or the resulting series of changes is the more fit for analysis. Ways to make this decision will be discussed. Time series data are often seasonal and methods for modelling seasonal time series will be covered, again with examples. . If time permits, unobserved component models will be briefly introduced.
Intended Audience: Intermediate SAS users, intermediate to advanced statistical knowledge
Tools Discussed: SAS/ETS
Prerequisite: Students should have some basic knowledge of ARIMA models. The examples should be of interest to anyone planning to analyze data taken over time.
Class Outline:
- Overview
- Autoregressive Models
- Model checking
- Prediction Intervals
- Moving Average Models
- Model Selection – AIC
- Stationarity – Unit Roots
- Determining Lag Differences for Unit Root Tests
- Models with Inputs (PROC AUTOREG)
- Detecting Outliers
- Seasonal Models
- (optional) Nonlinear Trends
- (optional) St. Petersburg Visitor Example
- (optional) Seasonal Unit Roots