Free Trial

Safari Books Online is a digital library providing on-demand subscription access to thousands of learning resources.

Share this Page URL

Chapter 19. Exponential Smoothing Models... > 19.2 Detail Example: 10-Year Treasur... - Pg. 543

Chapter 19: Exponential Smoothing Models for Time Series Data 543 Government agencies regularly make forecasts using historical data collected at fixed time intervals. Such government forecasts include daily volume of airline travel, monthly housing starts, monthly Consumer Price Index (CPI), and daily high, low, and average temperatures in a city. In the study of the stock market or business cycles, the historical patterns of economic indicators are of interest. The purpose of such forecasts is to be prepared for the future. Forecasting is at its best when a regular pattern underlies the process that needs forecasting. It is at its worst when one-of-a-kind events, such as catastrophes, interfere with that pattern. Forecasts might be off target. In a good forecasting system, off-target forecasts are investigated and taken as occasions to improve the system. Data Requirements for Exponential Smoothing The JMP Times Series platform is designed to analyze time series data. Exponential smoothing is a univariate method, because it has only a single continuous Y-variable. The minimum data for exponential smoothing consist of a single column of continuous Y-values, ordered from the oldest (first row) to the most recent (last row). One important assumption for exponential smoothing in JMP is that the time intervals between observations are equal and that there are no missing observations in the series. Columns containing identifiers, such as dates, are typically added to the data file. It is possible to analyze multivariate time series with more than one Y-variable, but this is beyond the scope of this book. 19.2 Detail Example: 10-Year Treasury Note Closing