Course Outline

Lecture 1. Get to know your data

a. Overview of forecasting: how do you predict the future? What kind of accuracy is possible?
b. Where to obtain data; data sources at Fuqua and on the web
c. How to move data around: useful things you can do with your word processor and spreadsheet
d. What to look for in data: seasonality, inflation, trends, cycles, etc.
e. How to transform data to reveal its structure; deflation, logging, seasonal adjustment
f. Illustration of basic operations in Statgraphics

Reading assignment: Statgraphics Tutorial Introduction, H&R Chapters 1 and 3 (note: ignore details of how to use Minitab, SAS, and Forecast Plus throughout the book); also glance at pages 320-322 (how to use price indices) and pages 338-342 (how to compute seasonal indices), we'll return to these in lecture 3

Additional lecture notes:

Famous forecasting quotes
How to shovel data around
Get to know your data
Inflation adjustment (deflation)
Seasonal adjustment
Stationarity and differencing
The logarithm transformation

Lecture 2. Introduction to forecasting

a. Forecasting a stationary series: the "mean" model
b. Forecasting a nonstationary series I: the trend line model
c. Forecasting a nonstationary series II: the random walk ("naive") model
d. How to identify a random walk: differencing and autocorrelation analysis
e. Geometric random walk: the basic stock price model
f. Three types of forecasts: estimation period, validation period, and long-term extrapolation
g. How to evaluate forecast errors and compare models

Reading: Skim Chapter 2 and read Chapter 4, plus the first few pages of Chapter 5 (naive models) in H&R
Additional lecture notes:

Linear trend
Random walk
Random walk with growth
Geometric random walk
Three types of forecasts: estimation period, validation period, and the future

Lecture 3. Modeling of seasonality

HW#1 due
a. General considerations in working with seasonal data: causes of seasonality, stability of seasonal patterns
b. Seasonal random walk; and seasonal random trend models
c. Seasonal adjustment by the ratio-to-moving-average method
d. Additive versus multiplicative seasonal adjustment
e. Adjustments for holidays and trading days
f. Trend/cycle decomposition of time series

Reading: Chapter 8 (decomposition of time series, calculation of seasonal indices)

Addtional notes:
Seasonal differencing
Seasonal random walk
Seasonal random trend
Lecture 3 slides: working with seasonal data (revised Sep. 11)
Data set for assignment #2

Lecture 4. Averaging and smoothing models

a. Simple moving average model
b. Exponential smoothing model
c. Combination of smoothing and seasonal adjustment
d. Robust models for noisy data
e. Massively parallel forecasting

Reading: Remainder of Chapter 5 (moving averages and smoothing methods)

Additional notes:
Averaging and exponential smoothing models
Spreadsheet implementation of seasonal adjustment and exponential smoothing (revised 9/17)

Lecture 5. Regression to mediocrity

HW#2 due
a. Indiana Jones and the temple of R-squared
b. Correlation matrices, autocorrelation and cross-correlation functions
c. Fitting simple regression models; interpreting output

Reading: Chapter 6

Additional notes:
Introduction to regression analysis
Example of regression analysis: predicting auto sales from personal income

Lecture 6. Time series regression models

a. Fitting time series regression models

Reading: Chapter 7

Additional notes:
Fitting time series regression models
What to look for in regression output

Lecture 7. Regression continued

HW#3 due
a. What's a good value for R-squared?
b. Not-so-simple regression models

Reading: Chapter 9

Additional notes:
What's a good value for R-squared?
Not-so-simple regression models

Lecture 8. Regression wrapup

a. The four horsemen of linear regression
b. What's the bottom line?
c. Stepwise and all-subsets regression
d. Business cycle indicators

Additional notes:
Testing the assumptions of linear regression
What's the bottom line? How to compare models
Stepwise and all-possible-regressions
Business Cycle Indicators

Lecture 9. Introduction to ARIMA models

a. Naive + Autoregressive + Exponential Smoothing = ARIMA
b. Using ACF and PACF plots to determine the "signature" of a time series
c. Fitting non-seasonal ARIMA models
d. The spectrum of ARIMA models

Reading: Chapter 10

Additional notes (revised 9/30):
Introduction to ARIMA: nonseasonal models
Identifying the order of differencing
Identifying the orders of AR or MA terms
Rules for identifying ARIMA models

Lecture 10. ARIMA continued; seasonal models

HW#4 due
a. Identification of seasonal models
b. Examples of seasonal model-fitting
c. Spreadsheet implementation

Additional notes:
Estimation of ARIMA models
Seasonal ARIMA models

Lecture 11. ARIMA wrapup, automation

Designated project" option (there is no HW#5)
a. Combination of ARIMA and regression models
b. Recap of steps in choosing a forecasting model
c. Automatic forecasting software

Additional notes:
ARIMA models with regressors
Steps in choosing a forecasting model
Forecasting flow chart
Automatic forecasting software

Lecture 12. Politics and ethics

a. Political and ethical issues in forecasting
b. Review of models: what to use and when

Reading: Chapter 11
Political and ethical issues in forecasting
How to avoid trouble
Data transformations and forecasting models: what to use and when
Notes concerning the final project (transforming, untransforming, comparing models, etc.)

Wednesday, October 16: Final projects are due at 5pm (the final exam is cancelled for lack of interest)

Last updated October 11, 1996. Always under construction.