SAS Institute Alert Note SN-008764

From: SAS Technical Support (tsdlist@unx.sas.com)
Date: Tue Nov 26 2002 - 14:11:46 EST

  • Next message: SAS Technical Support: "SAS Note V6-SYS.PROC-H023"

    SN-008764 ***Alert Note***

    PROC GLM may not compute the correct error sums of squares when solving
    the Expected Mean Squares for RANDOM effects

    Product: SAS/STAT
    Component: Multiple procedures
    Priority: ALERT

      If the TEST option is requested on the RANDOM statement and
      if the reported ERROR term for a random effect test is
      "almost all" MS(ERROR), then PROC GLM and PROC ANOVA will incorrectly
      report the ERROR DF, ERROR SS, ERROR MS, F-Value, PR > F under the Tests
      of Hypothesis for Mixed Model Analysis of Variance table.

      The exact criterion for "almost all" is if

         sum_i (l(i) * MS(i))**2 / DF(i) < 1e-8

      where: l(i) is the coefficient of the mean square of the non-error
                  components
             MS(i) is the mean square for the ith component, and
             DF(i) is the degrees of freedom for the ith component

      In the example below there is only one non-error component:

         Error: 0.0001*MS(A*B) + 0.9999*MS(Error)

      with

         Source DF Type III SS Mean Square F Value Pr > F

          A*B 11 0.01666472 0.00151497 1.21 0.3138

      so

               sum_i (l[i] * MS(i))**2 / DF(i)

            ~ (0.0001 * 0.00151497)**2 / 11 < 1e-8.

      There is no circumvention to this problem in PROC GLM or PROC ANOVA.

      However, you can use PROC MIXED with the METHOD=TYPE3 option on the
      PROC statement to obtain the correct results. In general, all RANDOM
      effects will go on the RANDOM statement (and not on the MODEL statement)
      in PROC MIXED.

      A Technical Support hot fix for Release 8.2 TSLEVEL TS2M0 for this
      issue is available at:

      http://www.sas.com/techsup/download/hotfix/82_sbcs_prod_list.html#008764

      For customers running SAS with Asian Language Support (DBCS), this
      hot fix should be downloaded from:

      http://www.sas.com/techsup/download/hotfix/82_dbcs_prod_list.html#008764

    Keywords:
      glm anova ems error expected mean square df satterthwaite degree freedom
      wrong mixed model anova f value statistic stat incorr
      proc glm may not compute the correct error sums of squares when solving
      the expected mean squares for random effects
      008764 8764

    SAS Note Revised On: Wed, 20 Nov 2002

    System Version Reported Version Fixed

    AIX 64 bit 8.2 TS2M0
    AIX 64 bit 9 TSM0
    AIX/6000 8 TSM0
    AIX/6000 8.1 TS1M0
    AIX/6000 8.2 TS2M0
    VM/ESA (CMS) 8 TSM0
    VM/ESA (CMS) 8.1 TS1M0
    VM/ESA (CMS) 8.2 TS2M0
    Compaq Tru64 UNIX 8 TSM0
    Compaq Tru64 UNIX 8.1 TS1M0
    Compaq Tru64 UNIX 8.2 TS2M0
    Compaq Tru64 UNIX 9 TSM0
    HP-UX Operating Systems 8 TSM0
    HP-UX Operating Systems 8.1 TS1M0
    HP-UX Operating Systems 8.2 TS2M0
    HP-UX Itanium 9 TSM0
    HP-UX 64 bit 8.2 TS2M0
    HP-UX 64 bit 9 TSM0
    Intel ABI 8 TSM0
    Intel ABI 8.1 TS1M0
    Intel ABI 8.2 TS2M0
    Intel Itanium Processor Family(IPF) 9 TSM0
    Linux 8.2 TS2M0
    Linux 9 TSM0
    MIPS ABI 8 TSM0
    MIPS ABI 8.1 TS1M0
    MIPS ABI 8.2 TS2M0
    OS/390 (MVS) 8 TSM0
    OS/390 (MVS) 8.1 TS1M0
    OS/390 (MVS) 8.2 TS2M0
    OS/390 (MVS) 9 TSM0
    IBM OS/2 8 TSM0
    IBM OS/2 8.1 TS1M0
    IBM OS/2 8.2 TS2M0
    Solaris 8 TSM0
    Solaris 8.1 TS1M0
    Solaris 8.2 TS2M0
    Solaris 64 bit 8.2 TS2M0
    Solaris 64 bit 9 TSM0
    OpenVMS VAX 8 TSM0
    OpenVMS VAX 8.1 TS1M0
    OpenVMS VAX 8.2 TS2M0
    OpenVMS Alpha 8 TSM0
    OpenVMS Alpha 8.1 TS1M0
    OpenVMS Alpha 8.2 TS2M0
    OpenVMS Alpha 9 TSM0
    Windows NT 8 TSM0
    Windows NT 8.1 TS1M0
    Windows NT 8.2 TS2M0
    Windows NT 9 TSM0
    Windows 2000 Datacenter Server 8 TSM0
    Windows 2000 Datacenter Server 8.1 TS1M0
    Windows 2000 Datacenter Server 8.2 TS2M0
    Windows 2000 Datacenter Server 9 TSM0
    Windows 2000 Professional 8 TSM0
    Windows 2000 Professional 8.1 TS1M0
    Windows 2000 Professional 8.2 TS2M0
    Windows 2000 Professional 9 TSM0
    Windows 2000 Server and Advanced Server 8 TSM0
    Windows 2000 Server and Advanced Server 8.1 TS1M0
    Windows 2000 Server and Advanced Server 8.2 TS2M0
    Windows 2000 Server and Advanced Server 9 TSM0
    Windows 95 8 TSM0
    Windows 95 8.1 TS1M0
    Windows 95 8.2 TS2M0
    Windows 98 8 TSM0
    Windows 98 8.1 TS1M0
    Windows 98 8.2 TS2M0
    Windows ME 8.2 TS2M0
    Windows XP 8.2 TS2M0
    Windows XP 9 TSM0
    Windows 64 bit 9 TSM0

    No Fixes Available

    +-------------------------------------------------------------------------
    + NOTE: To unsubscribe you must send mail to: LISTSERV@VM.SAS.COM with
    + "SIGNOFF tsnews-l"
    + as the only text in the body of the message (without the double quotes).
    + If you have any questions please send them to TSDLIST@SAS.COM
    +-------------------------------------------------------------------------



    This archive was generated by hypermail 2b29 : Tue Nov 26 2002 - 14:13:22 EST