SAS for Forecasting Time Series, Third Edition, is designed and written for those practitioners who are generally familiar with SAS, basic inferential statistics and with applications. While the book is not written for the theoretical statistician, it is written in such a manner as to be understood and used by those who are utilizing statistics to extract meaningful information from masses of time series data.

The book takes the reader through the various regression topics from simple regression (PROC REG) where, generally, there is no issue with invalid assumptions on through the autoregressive moving average model (PROC ARIMA) where autocovariances can be used to develop an appropriate model. Additional procedures addressed include PROC VARMAX, PROC SPECTRA, exponential smoothing and many other related topics and procedures. The recently developed SAS FORECAST STUDIO is used to present and explain many of the examples. Of particular interest is PROC X13, the SAS Procedure developed from the seasonal adjustment program of the U.S. Census Bureau.

Throughout the book real-life examples are used to illustrate and develop the applications of the procedures. The book, in many respects, offers an encyclopedic layout of time series applications. Data for examples and the full SAS programming is available on-line and is a valuable resource.

Kenneth L. Koonce, Professor Emeritus, Experimental Statistics
Louisiana State University

The authors are to be congratulated on their fine piece of work. The third edition of SAS^® for Forecasting Time Series contains four new chapters as well as updates to all of the chapters in the previous edition. The four new chapters cover the following topics:

• Exponential smoothing
• Unobserved components and state space models
• Adjustment for seasonality with PROC X13
• SAS Forecast Studio.

The new chapter on exponential smoothing show readers how to fit and interpret exponential smoothing for stationary data, trending data and seasonal data. The advantages of exponential smoothing are also discussed.

Unobserved components models are discussed in detail for both seasonal and nonseasonal models. The description of unobserved components models is expanded include an introduction to the broader and more flexible class of models known as state space models.

The new chapter on adjustment for seasonality with PROC X13 describes in detail the X-13 ARIMA-SEATS seasonal adjustment procedure developed by the U.S. Census Bureau for quarterly and monthly time series. The power of the X13 procedure is illustrated by applying it 918 Bureau of Labor Statistics monthly time series for the number of employees in different industries.

The fourth, and final, new chapter provides an introductory to tutorial to SAS Forecast Studio, a set of procedures that automatically fit univariate time series models to data. This material describes how to create a project in Forecast Studio along with all the modeling choices the user has at their disposal. As well as automatically detecting outliers and level shifts, it is possible to create custom events to model know outliers, such as specific holidays.

The SAS code for all the examples considered in the book is available online, as are the example data sets, making this book a highly valuable resource.

I highly recommend this book to anyone interested in using SAS to fit time series models. I expect to consult the third edition even more than I consulted the previous two editions.

Professor Simon Sheather, Interim Director of the Texas A&M Institute of Data Science
Academic Director of MS (Analytics) and Online Programs
Department of Statistics
Texas A&M University