Problem Note 62078: PROC SSM might return incorrect parameter estimates when fitting a model with an ARIMA trend type
If a model with an ARIMA trend type is estimated in PROC SSM by specifying the P=, Q=, SP=, or SQ= options in the TREND statement, the resulting parameter estimates and associated statistics might be incorrect. The incorrect results can be encountered when one or more of the P=, Q=, SP=, or SQ= options have a specified value that is greater than 1. Here are some examples:
trend trendspec(arima(p=3));
trend trendspec(arima(q=2));
trend trendspec(arima(p=2, d=1, q=1, sp=1, s=12));
trend trendspec(arima(q=3, sp=2, sq=1, s=4));
The incorrect parameter estimates and associated statistics can occur for models with orders greater than 1 in any of the nonseasonal or seasonal AR or MA parameters. This problem is not encountered when the nonseasonal or seasonal AR or MA terms specified in the TREND statement have an order that is at most 1. For example, the following specifications do not encounter the problem:
trend trendspec(arima(p=1, q=1, sp=1, s=12));
trend trendspec(arima(d=1, q=1, sq=1, s=4));
Other than specifying a model for the trend with ARMA orders that are at most 1, there is no circumvention for the problem.
Click the Hot Fix tab in this note to access the hot fix for this issue.
Operating System and Release Information
SAS System | SAS/ETS | z/OS | 12.3 | | 9.4 TS1M0 | |
Microsoft® Windows® for x64 | 12.3 | | 9.4 TS1M0 | |
Microsoft Windows 8 Enterprise x64 | 12.3 | | 9.4 TS1M0 | |
Solaris for x64 | 12.3 | | 9.4 TS1M0 | |
Microsoft Windows 8 Pro x64 | 12.3 | | 9.4 TS1M0 | |
Microsoft Windows 8.1 Enterprise 32-bit | 12.3 | | 9.4 TS1M0 | |
Microsoft Windows 8.1 Enterprise x64 | 12.3 | | 9.4 TS1M0 | |
Microsoft Windows 8.1 Pro 32-bit | 12.3 | | 9.4 TS1M0 | |
Microsoft Windows 8.1 Pro x64 | 12.3 | | 9.4 TS1M0 | |
Microsoft Windows 10 | 12.3 | | 9.4 TS1M0 | |
Microsoft Windows Server 2008 R2 | 12.3 | | 9.4 TS1M0 | |
Microsoft Windows Server 2008 for x64 | 12.3 | | 9.4 TS1M0 | |
Microsoft Windows Server 2012 Datacenter | 12.3 | | 9.4 TS1M0 | |
Microsoft Windows Server 2012 R2 Datacenter | 12.3 | | 9.4 TS1M0 | |
Microsoft Windows Server 2012 R2 Std | 12.3 | | 9.4 TS1M0 | |
Microsoft Windows Server 2012 Std | 12.3 | | 9.4 TS1M0 | |
Windows 7 Enterprise x64 | 12.3 | | 9.4 TS1M0 | |
Windows 7 Professional x64 | 12.3 | | 9.4 TS1M0 | |
64-bit Enabled AIX | 12.3 | | 9.4 TS1M0 | |
64-bit Enabled Solaris | 12.3 | | 9.4 TS1M0 | |
HP-UX IPF | 12.3 | | 9.4 TS1M0 | |
Linux for x64 | 12.3 | | 9.4 TS1M0 | |
*
For software releases that are not yet generally available, the Fixed
Release is the software release in which the problem is planned to be
fixed.
If a TREND statement with an ARIMA type is specified with P=, Q=, SP=, or SQ= greater than 1, PROC SSM might return incorrect parameter estimates and associated statistics.
Type: | Problem Note |
Priority: | high |
Topic: | Analytics ==> Econometrics Analytics ==> Forecasting Analytics ==> Time Series Analysis SAS Reference ==> Procedures ==> SSM
|
Date Modified: | 2018-04-11 10:37:14 |
Date Created: | 2018-04-03 16:19:50 |