Comparison of Forecasting Approaches Using Proc Reg and Proc Arima vs. SAS Time Series Forecasting System

• Activity Number: 248
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Computing
• Author(s): Martin Selzer*+
• Companies: Chatham Decision Sciences
• Keywords: SAS Macros ; Times Series Forecasting ; Regression ; Arima

Abstract

This paper describes a SAS macro based program using Proc Reg and Proc Arima developed for forecasting. The program is compared to a separate forecasting process using the SAS Time Series Forecasting System (SASTSFS). Advantages of using SAS macros to automate outcomes are discussed. Forecasts are prepared with the macro based program and compared to those produced by other programs and SASTSFS. Results show that the new program gives results quite near those generated by SASTSFS. While no effort was made to compare the accuracy of the two approaches against actual results, the mathematical formulation of the statistical models used in the macro based program suggest the program would provide more accurate outcomes. The macro based approach has fewer manual interventions and produces forecasts more quickly and less prone to error.