Now Playing: Forecasting and Time Series (2:57)
In NCSS, there are a variety of forecasting procedures for time series data. The use of each requires quite a bit of understanding of the background theory and methods. The purpose of this video is not to teach the details of the methods, but instead to showcase a couple of example procedure runs. Many articles and text books dedicated to the particulars of these methods are available to the interested learner.
Suppose that a company has 12 years of monthly sales data. Corporate leaders would like to forecast sales for the next two years. Two forecasting methods are under consideration: exponential smoothing and Box-Jenkins.
First, the desired options are specified for the Exponential Smoothing procedure, and the procedure is run. The forecast plot shows two years of prediction that appear to follow the trends of the previous years very nicely, including the December spike in sales, and the January and February lull. The forecasted values are shown at the end of the Forecasts section.
Next, the options are specified for use of the Box-Jenkins methods, and that procedure is run. The resulting forecasting plot is similar, but there are additionally confidence bands for the forecasted values. Several other summary reports are shown, including the forecasted values for each month.
The pros and cons of each method are not discussed here, but hopefully the viewer has a taste of what is needed to run a forecasting procedure in NCSS.