This example uses the SeriesP.jmp sample data table to show how to perform a time series analysis. You first create a new column that is appropriate for the Time ID.
1.
|
The SeriesP.jmp data table contains a Year column and a Quarter column to identify the time period during which the responses were observed. However, the Time Series platform requires one column with unique, equally spaced time points to label the X axis. If no Time ID is specified, then the row number is used to identify the time periods. To avoid this and make the report easier to interpret, you construct a Time ID column from Year and Quarter.
2.
|
3.
|
Select Column Properties > Formula.
|
4.
|
Select Year and then click the plus sign.
|
5.
|
6.
|
Click OK.
|
Figure 15.16 New Column
7.
|
Click OK.
|
1.
|
Select Analyze > Specialized Modeling > Time Series.
|
2.
|
3.
|
4.
|
Click OK.
|
Figure 15.17 Time Series Report for SeriesP.jmp
5.
|
Figure 15.18 Difference Report for SeriesP.jmp
7.
|
Click the Time Series GDP red triangle and select Smoothing Model > Linear Exponential Smoothing.
|
8.
|
Click Estimate.
|
9.
|
Click the Time Series GDP red triangle and select ARIMA Model Group. This enables you to fit multiple ARIMA models for a range of values of (p,d,q)(P,D,Q).
|
–
|
Fix d, the differencing order, at 1 by setting the range from 1 to 1 because the differencing report showed lag-1 differencing was appropriate.
|
–
|
Set p, the autoregressive order, to range from 0 to 1 because the original series showed evidence of autocorrelation.
|
–
|
Set q, the moving average order, to range from 0 to 1.
|
–
|
Figure 15.19 ARIMA Model Group Specification
11.
|
Click Estimate.
|
Figure 15.20 Model Comparison Table
Figure 15.21 Model Report for ARIMA(0,1,0)