To begin a time series analysis, choose the Time Series command from the Analyze > Modeling submenu to display the Time Series Launch dialog (The Time Series Launch Dialog). This dialog allows you to specify the number of lags to use in computing the autocorrelations and partial autocorrelations. It also lets you specify the number of future periods to forecast using each model fitted to the data. After you select analysis variables and click OK on this dialog, a platform launches with plots and accompanying text reports for each of the time series (Y) variables you specified.
You assign columns for analysis with the dialog in The Time Series Launch Dialog. The selector list at the left of the dialog shows all columns in the current table. To cast a column into a role, select one or more columns in the column selector list and click a role button. Or, drag variables from the column selector list to one of the following role boxes:
for the x-axis, one variable used for labeling the time axis
for the y-axis, one or more time series variables.
If you use a X, Time ID variable, you can specify the time frequency by using the Time Frequency column property. The choices are Annual, Monthly, Weekly, Daily, Hourly, Minute, and Second. This helps JMP determine the spacing of the data when plotting the forecast values. If no frequency is specified, the data is treated as equally spaced numeric data.
To remove an unwanted variable from an assigned role, select it in the role box and click Remove. After assigning roles, click OK to see the analysis for each time series variable versus the time ID.
You set the number of lags for the autocorrelation and partial autocorrelation plots in the Autocorrelation Lags box. This is the maximum number of periods between points used in the computation of the correlations. It must be more than one but less than the number of rows. A commonly used rule of thumb for the maximum number of lags is n/4, where n is the number of observations. The Forecast Periods box allows you to set the number of periods into the future that the fitted models are forecast. By default, JMP uses 25 lags and 25 forecast periods.
The data for the next examples are in the table found in the Time Series sample data folder (Box and Jenkins 1976). The time series variable is Passengers and the Time ID is Time.
The Time Series Launch Dialog
The Time Series platform begins with a plot of each times series by the time ID, or row number if no time ID is specified (Time Series Plot of Seriesg (Airline Passenger) Data). The plot, like others in JMP, has features to resize the graph, highlight points with the cursor or brush tool, and label points. See Using JMP for a discussion of these features.
Time Series Plot of Seriesg (Airline Passenger) Data
If you open Time Series Basic Diagnostic Tables, graphs of the autocorrelation and partial autocorrelation (Autocorrelation (left) and Partial Correlation (right) Graphs) of the time series are shown.