Problem Note 51281: The Henze-Zirkler multivariate normality test computed by PROC MODEL is incorrect
 |  |  |  |
If the NORMAL option is specified in the FIT statement in the MODEL procedure, the results for the Henze-Zirkler T test are incorrect.
As a circumvention, you can use the Mardia Skewness test and the Mardia Kurtosis test to test for multivariate normality. These tests are also printed by the NORMAL option in the FIT statement. For more information, see the "Testing for Normality" section of the PROC MODEL documentation under "Details: Estimation by the MODEL Procedure."
Operating System and Release Information
| Product Family | Product | System | Product Release | SAS Release | ||
| Reported | Fixed* | Reported | Fixed* | |||
| SAS System | SAS/ETS | Linux | 9.3 | 13.1 | 9.4 TS1M1 | |
| HP-UX IPF | 9.3 | 13.1 | 9.4 TS1M1 | |||
| 64-bit Enabled Solaris | 9.3 | 13.1 | 9.4 TS1M1 | |||
| 64-bit Enabled HP-UX | 9.3 | 13.1 | 9.4 TS1M1 | |||
| 64-bit Enabled AIX | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Windows Vista for x64 | 9.3 | |||||
| Windows Vista | 9.3 | |||||
| Windows Millennium Edition (Me) | 9.3 | |||||
| Windows 7 Ultimate x64 | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Windows 7 Ultimate 32 bit | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Windows 7 Professional x64 | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Windows 7 Professional 32 bit | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Windows 7 Home Premium x64 | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Windows 7 Home Premium 32 bit | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Windows 7 Enterprise x64 | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Windows 7 Enterprise 32 bit | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Microsoft Windows XP Professional | 9.3 | |||||
| Microsoft Windows Server 2012 Std | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Microsoft Windows Server 2012 Datacenter | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Microsoft Windows Server 2008 for x64 | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Microsoft Windows Server 2008 R2 | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Microsoft Windows Server 2008 | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Microsoft Windows Server 2003 for x64 | 9.3 | |||||
| Microsoft Windows Server 2003 Standard Edition | 9.3 | |||||
| Microsoft Windows Server 2003 Enterprise Edition | 9.3 | |||||
| Microsoft Windows Server 2003 Datacenter Edition | 9.3 | |||||
| Microsoft Windows NT Workstation | 9.3 | |||||
| Microsoft Windows 2000 Professional | 9.3 | |||||
| Microsoft Windows 2000 Server | 9.3 | |||||
| Microsoft Windows 2000 Datacenter Server | 9.3 | |||||
| Microsoft Windows 2000 Advanced Server | 9.3 | |||||
| Microsoft Windows 95/98 | 9.3 | |||||
| Microsoft Windows 8 Pro x64 | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Microsoft Windows 8 Pro 32-bit | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Microsoft Windows 8 Enterprise x64 | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Microsoft Windows 8 Enterprise 32-bit | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Microsoft® Windows® for x64 | 9.3 | 13.1 | 9.4 TS1M1 | |||
| z/OS | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Linux for x64 | 9.3 | 13.1 | 9.4 TS1M1 | |||
| Solaris for x64 | 9.3 | 13.1 | 9.4 TS1M1 | |||
If you specify the NORMAL option in the FIT statement in RPOC MODEL, the Henze-Zirkler T test and p-value are incorrect. They should not be used to test for multivariate normality of the equation errors.
| Type: | Problem Note |
| Priority: | alert |
| Topic: | SAS Reference ==> Procedures ==> MODEL Analytics ==> Distribution Analysis Analytics ==> Econometrics Analytics ==> Multivariate Analysis Analytics ==> analytics |
| Date Modified: | 2017-01-18 15:02:01 |
| Date Created: | 2013-10-08 11:37:48 |