PROC TABULATE produces incorrect percentage values
PROC MEANS and PROC SUMMARY may output incorrect number of observations
Error: Could not find value/rid to delete on index x for y
MODIFY with KEY=/UNIQUE causes a REMOVE on a non-match record
PROC MEANS and PROC SUMMARY produce incorrect confidence limits
PROC FREQ with the BY and OUTPUT statements may delete WORK data sets
PROC REPORT may display incorrect computed values when scrolling
Missing values for results of CONTRAST and ESTIMATE statements
Incorrect Likelihood Printed With RANDOM /G and REPEATED Statements
Incorrect N and SD statistics for MEAN option on REPEATED statement
PROC LIFETEST prints survival estimates when none are defined
Incorrect results possible with WEIGHT statement and missing values
Results of best subset selection incorrect with more than 50 variables
Incorrect OBSTATS when input variables used in link or distribution
Incorrect results for PROC MIXED with RANDOM and REPEATED specification
Incorrect Predicted Values When Using RANDOM with GROUP=
Class Variable Crossed with Identity Variable Does Not Create Variables
Incorrect DF and VALUE/DF when DIST=BINOMIAL is used with FREQ statement
OUTROC= data set variables incorrect when OFFSET= option is used
Incorrect P-Values Can Be Returned with the EXACT Statement
V= and SUBJECT= on RANDOM Statement Will Produce Incorrect Results
Potentially Incorrect Zero Variance Components with PROC MIXED
REGWQ critical values may not be correct in MEANS statement
Incorrect p-values reported for PDIFF option in PROC GLM and PROC MIXED
NOTE: The REGWF option has been disabled ... use REGWQ option instead
Incorrect Results From CONTRAST/ESTIMATE With Multiple RANDOM Statements
Incorrect Predicted Values with Multiple RANDOMs and Non-Nested SUBJECTs
Variance Components Incorrect for FA(0) With RANDOM Statement
Variance Component for CSH Parameter Can Be Incorrect
RATIOS Column in Models with FA(0) Structures Incorrect
Incorrect Derivative When Sum Statement Used in PROC MODEL or PROC NLP
DUNCAN option on MEANS statement gives wrong critical values
Incorrect Cp and BIC values may be given on plot
PROC GLM Degrees of Freedom can be wrong if SS2 SS3 SS4 is specified
Incorrect labels in OUT= data set with CLASS= and MAXNCAT= or MAXCMPCT=
Possible incorrect plot in key cell of a comparative histogram
WALD test statistic is incorrect when estimating a system of equations
Lagrange Multiplier Test statistic is incorrect for 2SLS, 3SLS, GMM
Copy of "other" variables incorrect when missing ID values at ends
GMM incorrectly handles model with negative sign on intercept term
The value may change for a numeric column whose length is less than 8
Using _GOTO_ABSOLUTE_ROW_ may copy changes from the current row
Values set in INIT of the model SCL may not be saved
A varchar2 column in a view changes blanks to be null in a where clause
Data set corruption may result with Win95 client and APPC access method (SAS/CONNECT)
Graphical Variance report may misrepresent the data
Data set corruption may result with Win95 client and APPC access method (SAS/SHARE)
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-PARETO-B392 *** ALERT NOTE *** PRODUCT: QC
PROCEDURE: PARETO
TITLE: Incorrect labels in OUT= data set with CLASS= and MAXNCAT= or MAXCMPCT=
KEYS: OUT= MAXNCAT MAXCMPCT CLASS PROCESS CATEGORIES WRONG ORDER OUTPUT DATA
SET CAUSE COUNT _COUNT_ INCORRSTAT INCORROUT
The OUT= data set in PROC PARETO will contain incorrect values for the
process variable when the CLASS= option is specified and either the
MAXNCAT= or MAXCMPCT= option is specified on the VBAR statement. All
of the numerical results in the output data set are correct but should
not be used since the associated process variable labels are wrong.
Note that if the options MAXNCAT= or MAXCMPCT= do not actually limit
the number of categories appearing on the chart, then the corresponding
OUT= data set will be correct.
There is no circumvention for the problem of incorrect labels.
+------REPORTED------+
CMS 6.07 TS301
MVS 6.07 TS301
VMS 6.07 TS301
OS/2 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 TS027
CONVEX 6.09 TS027
HP800 6.09 TS027
Solaris 6.09 TS027
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
WIN/NT 6.09 TS027
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-TABULATE-B424 *** ALERT NOTE *** PRODUCT: BASE
PROCEDURE: TABULATE
TITLE: PROC TABULATE produces incorrect percentage values
KEYS: PROC TABULATE PRODUCES INCORRECT PCTSUM VALUES INCORRSTAT PCTN ORDER
CALCULATE
In some circumstances PROC TABULATE may calculate the PCTSUM or PCTN
statistics incorrectly. The problem is related to the positioning of
other columns in the table. If multiple percentage statistics are
requested for the same crossing of variables, but these expressions are
not adjacent to one another in the TABLE statement, the resulting
statistics will be incorrect. For example, the following TABLE
statement produces incorrect values:
TABLE A*B*PCTSUM A*C*PCTSUM A*B*PCTN;
Notice that there are two expressions that both contain A and B crossed
with a percentage. Because these expressions are not next to one
another, the values in the A*C*PCTSUM and A*B*PCTN columns are not
computed correctly. To circumvent the problem, rearrange the columns of
the table so all crossings containing the same set of variables are
grouped together:
TABLE A*B*PCTSUM A*B*PCTN A*C*PCTSUM;
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404 6.08 TS450
MVS 6.08 TS404 6.08 TS450
OS/2 6.08 TS404 6.11 TS045
VMS 6.08 TS404 6.08 TS450
VSE 6.08 TS404
Windows 6.08 TS404 6.11 TS045
AIX/R 6.09 TS027 6.11 TS045
CONVEX 6.09 TS027 6.11 TS045
HP800 6.09 TS027 6.11 TS045
Solaris 6.09 TS027 6.11 TS045
ULTRIX 6.09 TS027 6.11 TS045
VMS/AXP 6.09 TS027 6.11 TS060
WIN/NT 6.09 TS027 6.11 TS045
IABI 6.10 TS018 6.11 TS045
MIPS ABI 6.10 TS018 6.11 TS045
Macintosh 6.10 TS037 6.11 TS045
OSF/AXP 6.10 TS018 6.11 TS045
OS/2 6.10 TS019 6.11 TS045
Windows 6.10 TS019 6.11 TS045
AIX/R 6.11 TS020 6.11 TS045
HP800 6.11 TS020 6.11 TS045
OS/2 6.11 TS020 6.11 TS045
Solaris 6.11 TS020 6.11 TS045
Windows 6.11 TS020 6.11 TS045
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MIXED-7641 OUTSTANDING PROBLEM PRODUCT: STAT
TITLE: Missing values for results of CONTRAST and ESTIMATE statements
KEYS: MISSING VALUES RESULTS CONTRAST ESTIMATE STATEMENTS INCORROUT
INCORRSTAT
The CONTRAST and ESTIMATE statements can produce incorrect results or
missing values when a variable name on the MODEL statement is wholly
contained in the first characters of a preceeding variable name. For
example,
MODEL Y = BLK BL;
will cause a CONTRAST or ESTIMATE statement involving the BL effect to
produce incorrect results or missing values, while the statement
MODEL Y = BL BLK;
will produce the correct results. The /E option on the CONTRAST or
ESTIMATE statement can be used to check the coefficients on each effect.
To circumvent the problem, change the order of the variables on the
MODEL statement
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.07 TS301
MVS 6.07 TS301
VMS 6.07 TS301
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404 6.10
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404 6.10
AIX/R 6.09 TS027 6.11
CONVEX 6.09 TS027
HP800 6.09 TS027 6.11
Solaris 6.09 TS027 6.11
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
WIN/NT 6.09 TS027 6.11
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MIXED-B580 *** ALERT NOTE *** PRODUCT: STAT
TITLE: Incorrect Likelihood Printed With RANDOM /G and REPEATED Statements
KEYS: INCORRECT LIKELIHOOD PRINTED RANDOM REPEATED STATISTICS PROC MIXED SBC
AIC LOGLIKELIHOOD
An incorrect likelihood value will be printed for models that use both
the REPEATED statement and the RANDOM statement with the /G option
when the estimated G matrix is singular. This incorrect value will
effect the AIC and the SBC as well. To circumvent the problem, remove
the /G option from the RANDOM statement and the correct value for
the likelihood statistic will be displayed.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 TS027
CONVEX 6.09 TS027
HP800 6.09 TS027
Solaris 6.09 TS027
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
WIN/NT 6.09 TS027
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
OS/2 6.10 TS019
Windows 6.10 TS019
AIX/R 6.11 TS020
HP800 6.11 TS020
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-GLM-B712 OUTSTANDING PROBLEM PRODUCT: STAT
PROCEDURE: GLM
TITLE: Incorrect N and SD statistics for MEAN option on REPEATED statement
KEYS: INCORRECT SD STATISTICS MEAN OPTION REPEATED N GLM STAT
INCORROUT INCORRSTAT STANDARD DEVIATION
When more than one repeated effect and the MEAN option is specified
on the REPEATED statement in PROC GLM, the following statistics are
incorrectly computed: N and SD . The MEAN value is computed correctly.
To circumvent the problem rearrange the data for a univariate analysis
and specify the within subject effects on the MEANS statement.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 TS027
CONVEX 6.09 TS027
HP800 6.09 TS027
Solaris 6.09 TS027
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
WIN/NT 6.09 TS027
Windows 6.10 TS019
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-ACCESS-B574 *** ALERT NOTE *** PRODUCT: ACCORA
PROCEDURE: ACCESS
TITLE: A varchar2 column in a view changes blanks to be null in a where clause
KEYS: VARCHAR2 COLUMN VIEW CHANGES BLANKS NULL WHERE ACCESS ORACLE 6.11
DESCRIPTOR UNIX BXD
If you create an access and view descriptor on an ORACLE table and
then compare a column of type VARCHAR2 in a where condition to 1 or
more blanks, the engine changes the blanks to a null value and
therefore gives incorrect data.
This problem occurs if a where clause as described above is used
in a datastep with a where or in a proc with a where clause.
This problem exists in the 6.11 platforms only. The 6.09 release
functions properly.
A work around for this is to include the where condition in
the view descriptor when you are initially creating it in
proc access.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
AIX/R 6.11 TS020
HP800 6.11 TS020
Solaris 6.11 TS020
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-SYS.PROC-B550 *** ALERT NOTE *** PRODUCT: BASE
PROCEDURE: PROC
TITLE: PROC MEANS and PROC SUMMARY may output incorrect number of observations
KEYS: PROC MEANS SUMMARY OUTPUT INCORRECT NUMBER OBSERVATIONS
When PROC MEANS or PROC SUMMARY have insufficient memory to process all
classification levels, they may swap some of the processing to disk.
When this occurs a note to that effect prints in the SAS log. However,
in some instances it has been discovered that when a BY statement is
used and there is insufficient memory the procedure may not write the
correct number of observations into the output data set. To insure that
the output data set is complete, increase the MEMSIZE= SAS system option
such that the procedure can complete all processing in memory.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404 6.09 TS450
MVS 6.08 TS404 6.09 TS450
OS/2 6.08 TS404
VMS 6.08 TS404 6.09 TS450
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 TS027
CONVEX 6.09 TS027
HP800 6.09 TS027
Solaris 6.09 TS027
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
WIN/NT 6.09 TS027
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018 6.11 TS040
OS/2 6.10 TS019
Windows 6.10 TS019
AIX/R 6.11 TS020
HP800 6.11 TS020
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-AF-B737 *** ALERT NOTE *** PRODUCT: AF
PROCEDURE: AF
TITLE: The value may change for a numeric column whose length is less than 8
KEYS: DATA FORM TABLE DATAFORM DATATABLE FRAME MODEL SCL NUMERIC VARIABLE
LENGTH LESS THAN EIGHT EXPONENTIAL VALUE INIT SCROLL UPDATE CHANGE
The value of a numeric column whose length is less than 8 may be updated
incorrectly in a Data Form or Data Table object within a SAS/AF FRAME
entry. The value for this column may change as you simply scroll through
the rows in edit mode.
To replicate this problem, you would need to have Data Form or Data
Table model SCL code that includes an INIT section and references
the name of the column somewhere in the SCL code.
To circumvent the problem, change the length of numeric variables
to 8 when they will be linked to columns in a Data Form or Data Table
object.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
OS/2 6.11 TS020
Windows 6.11 TS020
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-LIFETEST-B511 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: LIFETEST
TITLE: PROC LIFETEST prints survival estimates when none are defined
KEYS: PROC LIFETEST SURVIVAL ESTIMATE CENSOR INCORROUT INCORRSTAT
If you are using the TIMELIST= option to specify times greater
than the greatest time in the DATA= data set, and if the greatest
time in the DATA= data set is a censored observation, PROC
LIFETEST will print the survival estimates that correspond to the
last event time, rather than missing values.
Since the survival estimates are not defined for time values
greater than the last observed time when the last observed time
is censored, the estimates should appear as missing values in the
output.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
AIX/R 6.11 TS020
HP800 6.11 TS020
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MIXED-B672 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: MIXED
TITLE: Incorrect results possible with WEIGHT statement and missing values
KEYS: PROC MIXED WEIGHT MISSING VALUES STAT INCORRECT INCORROUT INCORRSTAT
MODEL FITTING INFORMATION RESULTS
If the response variable has missing values and a WEIGHT statement is
used, the statistics in the model fitting information table (REML Log
Likelihood, Akaike's Information Criterion, Schwarz's Bayesian
Criterion and -2 REML Log Likelihood) will be based on the total number
of observations, rather than the number of non-missing observations.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.07 TS301 6.09 TS450
MVS 6.07 TS301 6.09 TS450
VMS 6.07 TS301 6.09 TS450
CMS 6.08 TS404 6.09 TS450
MVS 6.08 TS404 6.09 TS450
OS/2 6.08 TS404 6.11 TS020
VMS 6.08 TS404 6.09 TS450
VSE 6.08 TS404
Windows 6.08 TS404 6.11 TS020
AIX/R 6.09 TS027 6.11 TS020
CONVEX 6.09 TS027
HP800 6.09 TS027
Solaris 6.09 TS027 6.11 TS020
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027 6.11 TS060
WIN/NT 6.09 TS027 6.11 TS040
IABI 6.10 TS018 6.11 TS040
MIPS ABI 6.10 TS018 6.11 TS040
Macintosh 6.10 TS037 6.11 TS060
OSF/AXP 6.10 TS018 6.11 TS060
OS/2 6.10 TS019 6.11 TS020
Windows 6.10 TS019 6.11 TS020
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-AF-B944 *** ALERT NOTE *** PRODUCT: AF
PROCEDURE: AF
TITLE: Using _GOTO_ABSOLUTE_ROW_ may copy changes from the current row
KEYS: GOTO ABSOLUTE ROW COPY CHANGES PREVIOUS CURRENT
DATAFORM DATATABLE FORM TABLE DATA VALUES CHANGE OVERWRITE
In a Data Table or Data Form object within a SAS/AF FRAME entry, if you
make changes to the current row and then move to a different row using
the _GOTO_ABSOLUTE_ROW_ method, the changes from the previous row may be
copied to the row to which you moved instead of the previous row.
To circumvent the problem, execute _UPDATE_ROW_ and _UNLOCK_ROW_
before _GOTO_ABSOLUTE_ROW_.
Within the Data Form object, you may see the same problem if you
make a change to the current row and then scroll to a different row
by issuing the number of the row as a command.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
AIX/R 6.11 (???)
HP800 6.11 (???)
OS/2 6.11 (???)
Solaris 6.11 (???)
Windows 6.11 (???)
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-SYS.PROC-B849 *** ALERT NOTE *** PRODUCT: STAT
TITLE: Results of best subset selection incorrect with more than 50 variables
KEYS: REG LOGISTIC PHREG SELECTION SELECTION= SCORE RSQUARE ADJRSQ CP
INCORRSTAT INCORROUT ORDER MISORDER
Incorrect output may be generated in the REG, LOGISTIC or PHREG
procedures when "best subset" model selection methods are used to select
among a large number of possible explanatory variables. In PROC
LOGISTIC or PROC PHREG, the problem can occur when the SELECTION=SCORE
option is specified. In PROC REG, it can occur when the SELECTION=
value is RSQUARE, ADJRSQ or CP. If there are more than 50 variables
specified in the MODEL statement, then the printed results will be
incorrect. Values of statistics printed with the models and/or the
order of the printed models can be incorrect. This problem can only be
avoided by using another model selection method (such as STEPWISE) or by
restricting the number of explanatory variables to 50 or fewer.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
OS/2 6.06
VMS 6.06
Apollo 6.07 TS101
AIX/R 6.07 TS101
AViiON 6.07 TS101
BULL 6.07 TS201
HP300 6.07 TS101
HP800 6.07 TS201
MIPS 6.07 TS101
NeXT 6.07 TS201
SGI 6.07 TS201
Siemens 6.07 TS201
Sun-3 6.07 TS201
Sun-4 6.07 TS201
ULTRIX 6.07 TS101
VMS 6.07 TS301
OS/2 6.08 TS404
VMS 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 (???)
CONVEX 6.09 (???)
HP800 6.09 (???)
Solaris 6.09 (???)
ULTRIX 6.09 (???)
VMS/AXP 6.09 (???)
WIN/NT 6.09 (???)
IABI 6.10 (???)
MIPS ABI 6.10 (???)
Macintosh 6.10 (???)
OSF/AXP 6.10 (???)
OS/2 6.10 (???)
Windows 6.10 (???)
AIX/R 6.11 (???)
HP800 6.11 (???)
OS/2 6.11 (???)
Solaris 6.11 (???)
Windows 6.11 (???)
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-GENMOD-B887 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: GENMOD
TITLE: Incorrect OBSTATS when input variables used in link or distribution
KEYS: INCORRSTAT INCORROUT OBSTATS INPUT DATA SET VARIABLE DEVIANCE VARIANCE
FWDLINK INVLINK
If one or more variables from the input data set is used in any of the
DEVIANCE, VARIANCE, FWDLINK, or INVLINK statements, then the results of
the OBSTATS option may be incorrect.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS415
MVS 6.08 TS415
VMS 6.08 TS420
VSE 6.08 TS415
AIX/R 6.09 (???) 6.12 (???)
HP800 6.09 (???) 6.12 (???)
Solaris 6.09 (???) 6.12 (???)
VMS/AXP 6.09 (???)
IABI 6.10 (???)
MIPS ABI 6.10 (???) 6.12 (???)
Macintosh 6.10 (???)
OSF/AXP 6.10 (???) 6.12 (???)
OS/2 6.10 (???) 6.12 (???)
Windows 6.10 (???) 6.12 (???)
AIX/R 6.11 (???) 6.12 (???)
HP800 6.11 (???) 6.12 (???)
OS/2 6.11 (???) 6.12 (???)
Solaris 6.11 (???) 6.12 (???)
Windows 6.11 (???) 6.12 (???)
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MIXED-B724 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: MIXED
TITLE: Incorrect results for PROC MIXED with RANDOM and REPEATED specification
KEYS: INCORRECT RESULTS PROC MIXED RANDOM REPEATED TYPE= UN DDFM SATTERTH
INCORRSTAT INCORROUT
The DDFM=SATTERTH option will produce incorrect degrees of freedom for
models involving a REPEATED statement. This error will affect the
p-values as well.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS420
MVS 6.08 TS420
VMS 6.08 TS420
VSE 6.08 TS420
AIX/R 6.09 (???) 6.12 (???)
CONVEX 6.09 (???)
HP800 6.09 (???) 6.12 (???)
Solaris 6.09 (???) 6.12 (???)
ULTRIX 6.09 (???)
VMS/AXP 6.09 (???)
IABI 6.10 (???) 6.12 (???)
MIPS ABI 6.10 (???) 6.12 (???)
Macintosh 6.10 (???)
OSF/AXP 6.10 (???) 6.12 (???)
OS/2 6.10 (???) 6.12 (???)
Windows 6.10 (???) 6.12 (???)
AIX/R 6.11 (???) 6.12 (???)
HP800 6.11 (???) 6.12 (???)
OS/2 6.11 (???) 6.12 (???)
Solaris 6.11 (???) 6.12 (???)
Windows 6.11 (???) 6.12 (???)
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MIXED-B925 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: MIXED
TITLE: Incorrect Predicted Values When Using RANDOM with GROUP=
KEYS: INCORRECT PREDICTED VALUES USING RANDOM GROUP REPEATED SUBJECT
PROC MIXED can produce incorrect missing values when a RANDOM
statement is used with the GROUP= option. Incorrect values can also
be produced with the use of a RANDOM and a REPEATED statement when
completely different (non-nested) SUBJECT= effects are used.
There is no circumvention for this problem.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 (???)
CONVEX 6.09 (???)
HP800 6.09 (???)
Solaris 6.09 (???)
ULTRIX 6.09 (???)
VMS/AXP 6.09 (???)
WIN/NT 6.09 (???)
IABI 6.10 (???)
MIPS ABI 6.10 (???)
Macintosh 6.10 (???)
OSF/AXP 6.10 (???)
OS/2 6.10 (???)
Windows 6.10 (???)
AIX/R 6.11 (???)
HP800 6.11 (???)
OS/2 6.11 (???)
Solaris 6.11 (???)
Windows 6.11 (???)
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
+-<<< R6.07 SAS Notes V6-TRANSREG-B934 >>>-+
TITLE:
Class Variable Crossed with Identity Variable Does Not Create Variables
PRODUCT: STAT DIVISION: STATISTICS SAS NOTE: B934
MODULE: TRANSREG SUPPORT: KUHFELD, W. PRIORITY: ALERT
STATUS: OUTSTANDING PROBLEM BUGID: D0094742 PORTABLE? Y
TEXT:
In PROC TRANSREG, when the CLASS variable is crossed with the IDENTITY
variable list, all variables are not always created. For example, in:
PROC TRANSREG DESIGN;
MODEL CLASS(X / ZERO=NONE) | IDE (Z1-Z5);
the variables Z$ and Z5 are not being crossed with the CLASS variables.
To circumvent this problem cross all variables as shown below:
PROC TRANSREG DESIGN;
MODEL CLASS (X/ZERO=NONE)
CLASS (X/ZERO=NONE) * IDE(Z1)
CLASS (X/ZERO=NONE) * IDE(Z2)
CLASS (X/ZERO=NONE) * IDE(Z3)
CLASS (X/ZERO=NONE) * IDE(Z4)
CLASS (X/ZERO-NONE) * IDE(Z5)
OUTPUT;
RUN;
KEYS:
TRANSREG CLASS IDENTITY MODEL
>
----REPORTED----+ +-----FIXED------+
SYSTEM RLSE LVL RLSE LVL DEV-FIX
>
AIX/R 6.11 020 6.12 020
HP800 6.11 020 6.12 020
OS/2 6.11 020 6.12 020
Solaris 6.11 020 6.12 020
Windows 6.11 020 6.12 020
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MODEL-B987 *** ALERT NOTE *** PRODUCT: ETS
TITLE: WALD test statistic is incorrect when estimating a system of equations
KEYS: WALD TEST STATISTIC INCORRECT ESTIMATING SYSTEM EQUATIONS INCORRSTAT
INCORROUT STATEMENT P-VALUE HYPOTHESIS SCALE
When a Wald statistic is requested on the TEST statement either by
default, or explicitly with the WALD or ALL options, and a system of
equations is being estimated, the WALD statistic and its associated
p-value are incorrect. This is regardless of the estimation method
specified. If the OUT= option is also used on the TEST statement,
then the incorrect results are output to the OUT= data set.
To circumvent the problem, request a likelihood ratio test with the
LR option on the TEST statement.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
AIX/R 6.11 (???)
HP800 6.11 (???)
IABI 6.11 (???)
MIPS ABI 6.11 (???)
OSF/AXP 6.11 (???)
OS/2 6.11 (???)
Solaris 6.11 (???)
Windows 6.11 (???)
WIN/NT 6.11 (???)
Win95 6.11 (???)
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MODEL-B973 *** ALERT NOTE *** PRODUCT: ETS
TITLE: Lagrange Multiplier Test statistic is incorrect for 2SLS, 3SLS, GMM
KEYS: LAGRANGE MULTIPLIER TEST STATISTIC INCORRECT 2SLS 3SLS GMM LM RAO
INCORROUT INCORRSTAT P-VALUE HYPOTHESIS SCALE
When a Lagrange Multiplier test is requested on the TEST statement
with either the LM or ALL options, and the estimation method specified
is 2SLS, 3SLS or GMM, the Lagrange Multiplier test statistic and its
associated p-value are incorrect. If the OUT= option is also used
on the TEST statement, then the incorrect results are output to the
OUT= data set.
To circumvent the problem, request a likelihood ratio test.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
AIX/R 6.11 (???)
HP800 6.11 (???)
IABI 6.11 (???)
MIPS ABI 6.11 (???)
OSF/AXP 6.11 (???)
OS/2 6.11 (???)
Solaris 6.11 (???)
Windows 6.11 (???)
WIN/NT 6.11 (???)
Win95 6.11 (???)
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-SYS.DATA-C016 *** ALERT NOTE *** PRODUCT: BASE
PROCEDURE: DATA
TITLE: Error: Could not find value/rid to delete on index x for y
KEYS: COULD FIND VALUE RID DELETE INDEX CORRUPT INCORRECT RESULT INVALID
UPDATE IN PLACE MODIFY FSEDIT FSVIEW APPEND
Any DATA or PROC step which performs updates-in-place by either deleting
or changing an existing observation on an indexed dataset may cause
index corruption. Adding new observations will not cause this problem.
The problem is most likely to occur on an index with a pagesize of 27K
or larger. It is possible, but unlikely, for the problem to occur if
the index pagesize is smaller then 27K. In addition, the likelihood of
encountering the problem increases as the total number of updates
(deletes or changes) increases during the life of the dataset. The
following SAS statements or procedures perform an update-in-place:
MODIFY, SCL programs, PROC FSEDIT, PROC FSVIEW, and PROC SQL. The
following error message may or may not be issued to indicate index
corruption:
ERROR: Could not find value/rid to delete on index x for file y
A PROC or DATA step retrieving by a corrupted index may produce
incorrect results. The incorrect results may not be obvious to
the user. The following SAS statements may retrieve by index:
WHERE, BY, MODIFY or SET with KEY= option.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404 6.09 TS450
MVS 6.08 TS404 6.09 TS450
OS/2 6.08 TS404
VMS 6.08 TS404 6.09 TS450
VSE 6.08 TS404
Windows 6.08 TS404 6.11 (???)
AIX/R 6.09 (???)
CONVEX 6.09 (???)
HP800 6.09 (???)
Solaris 6.09 (???)
ULTRIX 6.09 (???)
VMS/AXP 6.09 (???)
WIN/NT 6.09 (???) 6.11 (???)
IABI 6.10 (???) 6.11 (???)
MIPS ABI 6.10 (???) 6.11 (???)
Macintosh 6.10 (???)
OSF/AXP 6.10 (???) 6.11 (???)
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-SYS.DATA-C121 *** ALERT NOTE *** PRODUCT: BASE
PROCEDURE: DATA
TITLE: MODIFY with KEY=/UNIQUE causes a REMOVE on a non-match record
KEYS: KEYS: NON-MATCH MODIFY KEY= UNIQUE REMOVE INDEX _IORC_ _SOK
Using the MODIFY statement with the UNIQUE option to provide a
consecutive KEY= value whose matching records were deleted in prior
interations causes SAS to incorrectly REMOVE the next logical, non-match
record in the index. A subsequent read of a consecutive KEY= value on a
non-match record incorrectly produces an _IORC_ value of zero. The
following DATA step replicates the problem:
master tran
OBS X Y OBS X Y
1 1 3 1 1 2
2 2 4
3 3 5
data master ;
set tran ;
do i = 1 to 3 ;
modify master key = x / unique ;
if _iorc_ = %sysrc(_sok) then remove ;
else _error_ = 0 ;
end ;
run ;
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 (???)
CONVEX 6.09 (???)
HP800 6.09 (???)
Solaris 6.09 (???)
ULTRIX 6.09 (???)
VMS/AXP 6.09 (???)
WIN/NT 6.09 (???)
IABI 6.10 (???)
MIPS ABI 6.10 (???)
Macintosh 6.10 (???)
OSF/AXP 6.10 (???)
Win95 6.11 (???)
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-GENMOD-C144 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: GENMOD
TITLE: Incorrect DF and VALUE/DF when DIST=BINOMIAL is used with FREQ statement
KEYS: INCORRSTAT INCORROUT DISTRIBUTION BINOMIAL DIST=BIN FREQ DEGREES OF
FREEDOM DF VALUE/DF CRITERIA FOR ASSESSING GOODNESS FIT WHEN
When the DIST=BINOMIAL option is used in conjunction with the FREQ
statement, the DF and VALUE/DF columns in the Criteria for Assessing
Goodness of Fit table are incorrect. The values of the FREQ variable
are incorrectly ignored when computing degrees of freedom in binomial
models. Correct values can be obtained by replicating each observation
as many times as its FREQ value, and running PROC GENMOD without the
FREQ statement.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404 6.12 (???)
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404 6.12 (???)
AIX/R 6.09 (???) 6.12 (???)
CONVEX 6.09 (???)
HP800 6.09 (???) 6.12 (???)
Solaris 6.09 (???) 6.12 (???)
ULTRIX 6.09 (???)
VMS/AXP 6.09 (???)
WIN/NT 6.09 (???) 6.12 (???)
IABI 6.10 (???) 6.12 (???)
MIPS ABI 6.10 (???) 6.12 (???)
Macintosh 6.10 (???)
OSF/AXP 6.10 (???) 6.12 (???)
OS/2 6.10 (???) 6.12 (???)
Windows 6.10 (???) 6.12 (???)
AIX/R 6.11 (???) 6.12 (???)
HP800 6.11 (???) 6.12 (???)
IABI 6.11 (???) 6.12 (???)
MIPS ABI 6.11 (???) 6.12 (???)
OSF/AXP 6.11 (???) 6.12 (???)
OS/2 6.11 (???) 6.12 (???)
Solaris 6.11 (???) 6.12 (???)
Windows 6.11 (???) 6.12 (???)
WIN/NT 6.11 (???) 6.12 (???)
Win95 6.11 (???)
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-LOGISTIC-C111 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: LOGISTIC
TITLE: OUTROC= data set variables incorrect when OFFSET= option is used
KEYS: INCORRSTAT INCORROUT OFFSET OUTROC ROC PREDICTED PROBABILITIES RECEIVER
OPERATING CHARACTERISTIC CURVE OUTPUT DATA SET
If the OFFSET= and OUTROC= options are both specified, the values in the
OUTROC= data set will be incorrect. The offset variable is incorrectly
omitted from the computation of the predicted probabilities which are
then used to compute the variables in the OUTROC= data set. The data
set contents are correct if the OFFSET= option is not used.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
AIX/R 6.11 (???) 6.12 (???)
HP800 6.11 (???) 6.12 (???)
IABI 6.11 (???) 6.12 (???)
MIPS ABI 6.11 (???) 6.12 (???)
OSF/AXP 6.11 (???) 6.12 (???)
OS/2 6.11 (???) 6.12 (???)
Solaris 6.11 (???) 6.12 (???)
Windows 6.11 (???) 6.12 (???)
WIN/NT 6.11 (???) 6.12 (???)
Win95 6.11 (???) 6.12
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MEANS-C132 *** ALERT NOTE *** PRODUCT: BASE
PROCEDURE: MEANS
TITLE: PROC MEANS and PROC SUMMARY produce incorrect confidence limits
KEYS: PROC MEANS SUMMARY PRODUCES CLM LCLM UCLM CONFIDENCE LIMITS INCORRSTAT
COMPUTES CALCULATE WEIGHT
Beginning in release 6.07 PROC MEANS and PROC SUMMARY compute incorrect
confidence limits (CLM LCLM UCLM) in the presence of a WEIGHT statement.
Currently, no circumvention is available.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
Apollo 6.07 TS101
AIX/R 6.07 TS101 6.12 (???)
AViiON 6.07 TS101
BULL 6.07 TS201
CMS 6.07 TS301 6.09 TS450
HP300 6.07 TS101
HP800 6.07 TS201 6.12 (???)
MIPS 6.07 TS101 6.12 (???)
MVS 6.07 TS301 6.09 TS450
NeXT 6.07 TS201
SGI 6.07 TS201 6.12 (???)
Siemens 6.07 TS201
Sun-3 6.07 TS201
Sun-4 6.07 TS201 6.12 (???)
ULTRIX 6.07 TS101
VMS 6.07 TS301 6.09 TS450
CMS 6.08 TS404 6.09 TS450
MVS 6.08 TS404 6.09 TS450
OS/2 6.08 TS404 6.12 (???)
VMS 6.08 TS404 6.09 TS450
VSE 6.08 TS404
Windows 6.08 TS404 6.12 (???)
AIX/R 6.09 (???) 6.12 (???)
CONVEX 6.09 (???)
HP800 6.09 (???) 6.12 (???)
Solaris 6.09 (???) 6.12 (???)
ULTRIX 6.09 (???)
VMS/AXP 6.09 (???)
WIN/NT 6.09 (???) 6.12 (???)
IABI 6.10 (???) 6.12 (???)
MIPS ABI 6.10 (???) 6.12 (???)
Macintosh 6.10 (???)
OSF/AXP 6.10 (???) 6.12 (???)
OS/2 6.10 (???) 6.12 (???)
Windows 6.10 (???) 6.12 (???)
AIX/R 6.11 (???) 6.12 (???)
HP800 6.11 (???) 6.12 (???)
IABI 6.11 (???) 6.12 (???)
MIPS ABI 6.11 (???) 6.12 (???)
OSF/AXP 6.11 (???) 6.12 (???)
OS/2 6.11 (???) 6.12 (???)
Solaris 6.11 (???) 6.12 (???)
Windows 6.11 (???) 6.12 (???)
WIN/NT 6.11 (???) 6.12 (???)
Win95 6.11 (???) 6.12 (???)
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-FREQ-B877 *** ALERT NOTE *** PRODUCT: BASE
PROCEDURE: FREQ
TITLE: PROC FREQ with the BY and OUTPUT statements may delete WORK data sets
KEYS: PROC FREQ OUTPUT STATEMENTS DELETES DATA SETS DATA1 DATAN DELETE ERASE
WARNING
When PROC FREQ is executed with the BY and OUTPUT statements, and you
have created data set names using the DATA1, DATA2, ... DATAn
convention, PROC FREQ may delete one of these data sets. This deletion
occurs without any warning or error message.
To prevent the problem from happening, avoid creating temporary data
sets using this naming convention.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
AIX/R 6.11 (???) 6.12 (???)
HP800 6.11 (???) 6.12 (???)
OS/2 6.11 (???) 6.12 (???)
Solaris 6.11 (???) 6.12 (???)
Windows 6.11 (???) 6.11 (???)
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-NPAR1WAY-C257 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: NPAR1WAY
TITLE: Incorrect P-Values Can Be Returned with the EXACT Statement
KEYS: INCORRECT RETURNED EXACT P-VALUES NPAR1WAY PROC STAT
Under rare circumstances an incorrect P-value can be returned from the
EXACT statement. This behaviour will only occur in 2x2 cases where
the following is true:
Let the 4 cell counts for the 2x2 table be denoted as c(1,1), c(1,2),
c(2,1) and c(2,2). Define the column totals as c1=c(1,1)+c(2,1) and
c2=c(1,2)+c(2,2). Define the row totals r1 and r2 accordingly.
Define the average ranks for the colums as u1=(c1+1)/2 and
u2=c1 + (c2+1)/2. Let N be the total sample size. Then calculate
the following terms:
z = c(2,2) - 2*r2*c2/N
c = z - (r2+c2-N)
y = c(2,1)*u1 + c(2,2)*u2 - (N+1)*r2/2
The bug will occur only if y>0, c=0, and z>=0.
There is no circumvention for this problem.
+------REPORTED------+
SYSTEM RELEASE LEVEL
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
WIN/NT 6.11 TS040
Win95 6.11 TS040
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MIXED-C252 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: MIXED
TITLE: V= and SUBJECT= on RANDOM Statement Will Produce Incorrect Results
KEYS: V= SUBJECT= RANDOM INCORRECT RESULTS LOG LIKELIHOOD RESIDUAL PROC MIXED
STAT
The V= option used with the SUBJECT= option on the RANDOM statement
in PROC MIXED will cause the procedure to print incorrect values
for the log-likelihood statistics and for the residual variance
estimate.
To circumvent the problem, either drop the V= option or rewrite the
RANDOM statement without using the SUBJECT= effect.
+------REPORTED------+
SYSTEM RELEASE LEVEL
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 TS027
CMS 6.09 TS450
CONVEX 6.09 TS027
HP800 6.09 TS027
MVS 6.09 TS450
Solaris 6.09 TS027
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
VMS 6.09 TS450
WIN/NT 6.09 TS027
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
OS/2 6.10 TS019
Windows 6.10 TS019
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
WIN/NT 6.11 TS040
Win95 6.11 TS040
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MIXED-C292 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: MIXED
TITLE: Potentially Incorrect Zero Variance Components with PROC MIXED
KEYS: ZERO VARIANCE COMPONENTS PROC MIXED STAT REML MIVQUE0 INCORRSTAT
INCORROUT
PROC MIXED can potentially produce incorrect zero variance
component estimates with the default METHOD=REML. If PROC MIXED
computes the variance component of the effect to be between
1e-16 and 1e-6 at any time during the Newton-Raphson iteration process,
the procedure may treat that variance component as constrained to be
zero. This behavior is rare and will not happen for all models with
zero variance components. You can check the iteration history with the
ITDETAILS option in the PROC MIXED statement or the variance component
estimates can be checked with PROC VARCOMP using METHOD=REML to see if
you have encountered the problem.
This problem is generally caused by a poor selection of starting values.
To circumvent the problem, use the PARMS statement with the OLS option
to change the starting values or provide new nonzero starting values
with the PARMS statement.
+------REPORTED------+
SYSTEM RELEASE LEVEL
CMS 6.08 TS415
MVS 6.08 TS415
OS/2 6.08 TS407
VMS 6.08 TS415
VSE 6.08 TS415
Windows 6.08 TS407
AIX/R 6.09 TS039
CONVEX 6.09 TS039
HP800 6.09 TS039
Solaris 6.09 TS039
ULTRIX 6.09 TS039
VMS/AXP 6.09 TS039
WIN/NT 6.09 TS027
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
WIN/NT 6.11 TS040
Win95 6.11 TS040
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-SYS.PROC-C295 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: PROC
TITLE: REGWQ critical values may not be correct in MEANS statement
KEYS: REGWQ CRITICAL VALUES MEANS STATEMENT ANOVA GLM STAT INCORROUT
INCORRSTAT
The critical values for the REGWQ multiple comparison test in PROC
ANOVA or GLM should be monotone non-decreasing in the number of means.
Occasionally, they are not if the error degrees of freedom is
"relatively small." A reference that discusses this issue is
Multiple Comparisons: Theory and Methods, by Jason Hsu (1996).
The publisher is Chapman & Hall.
+------REPORTED------+
SYSTEM RELEASE LEVEL
Apollo 6.07 TS101
AIX/R 6.07 TS101
AViiON 6.07 TS101
BULL 6.07 TS201
CMS 6.07 TS301
HP300 6.07 TS101
HP800 6.07 TS201
MIPS 6.07 TS101
MVS 6.07 TS301
NeXT 6.07 TS201
SGI 6.07 TS201
Siemens 6.07 TS201
Sun-3 6.07 TS201
Sun-4 6.07 TS201
ULTRIX 6.07 TS101
VMS 6.07 TS301
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 TS027
CMS 6.09 TS450
CONVEX 6.09 TS027
HP800 6.09 TS027
MVS 6.09 TS450
Solaris 6.09 TS027
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
VMS 6.09 TS450
WIN/NT 6.09 TS027
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
OS/2 6.10 TS019
Windows 6.10 TS019
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
WIN/NT 6.11 TS040
Win95 6.11 TS040
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-SYS.PROC-C298 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: PROC
TITLE: Incorrect p-values reported for PDIFF option in PROC GLM and PROC MIXED
KEYS: INCORRECT VALUES REPORTED PDIFF OPTION PROC GLM MIXED ADJUST SIMULATE
DUNNETT INCORRSTAT LSMEANS INCORROUT
If the LSMEANS are very unequally correlated, then the p-values reported
in the PDIFF table with the ADJUST=SIMULATE or ADJUST=DUNNETT may
be incorrect. This behavior will only happen in very rare
circumstances in PROC GLM or PROC MIXED.
+------REPORTED------+
SYSTEM RELEASE LEVEL
CMS 6.08 TS420
MVS 6.08 TS420
VMS 6.08 TS420
VSE 6.08 TS420
AIX/R 6.09 TS027
CMS 6.09 TS450
CONVEX 6.09 TS042
HP800 6.09 TS042
MVS 6.09 TS450
Solaris 6.09 TS042
ULTRIX 6.09 TS042
VMS 6.09 TS450
WIN/NT 6.09 TS027
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
OS/2 6.10 TS019
Windows 6.10 TS019
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
WIN/NT 6.11 TS040
Win95 6.11 TS040
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-SYS.PROC-C294 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: PROC
TITLE: NOTE: The REGWF option has been disabled ... use REGWQ option instead
KEYS: REGWF REGWQ MEANS MULTIPLE COMPARISON ANOVA GLM STAT INCORROUT
INCORRSTAT NOTE DISABLED
The REGWF multiple comparison test was incorrectly implemented
in the MEANS statement.
The underlying problem was that our implementation incorrectly
assumed that only contiguous subsets for the groups ordered by
sample means needed to be tested for equality, as is the case with
REGWQ. In general, for REGWF all subsets of means must be tested
for equality.
+------REPORTED------+
SYSTEM RELEASE LEVEL
Apollo 6.07 TS101
AIX/R 6.07 TS101
AViiON 6.07 TS101
BULL 6.07 TS201
CMS 6.07 TS301
HP300 6.07 TS101
HP800 6.07 TS201
MIPS 6.07 TS101
MVS 6.07 TS301
NeXT 6.07 TS201
SGI 6.07 TS201
Siemens 6.07 TS201
Sun-3 6.07 TS201
Sun-4 6.07 TS201
ULTRIX 6.07 TS101
VMS 6.07 TS301
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 TS027
CMS 6.09 TS450
CONVEX 6.09 TS027
HP800 6.09 TS027
MVS 6.09 TS450
Solaris 6.09 TS027
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
VMS 6.09 TS450
WIN/NT 6.09 TS027
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
OS/2 6.10 TS019
Windows 6.10 TS019
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
WIN/NT 6.11 TS040
Win95 6.11 TS040
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-EXPAND-C367 *** ALERT NOTE *** PRODUCT: ETS
PROCEDURE: EXPAND
TITLE: Copy of "other" variables incorrect when missing ID values at ends
KEYS: COPY "OTHER" DATA VARIABLES INCORRECT MISSING VALUES ENDS INCORROUT
TRUNCATED OUTPUT ID DATE
When PROC EXPAND is used to interpolate missing values for some of the
variables on a data set, all variables not processed by a CONVERT
statement are copied from the input data set to the OUT= data set.
However, if the ID variable has missing values at the beginning or end
of the data set, then the range of observations output to the OUT=
data set should be truncated. The copy of the input observations for
the variables not processed by a CONVERT statement does not take this
into account, therefore, when missing values occur at the beginning or
end of the file, then wrong values are copied to the OUT= data set.
To circumvent the problem, use a WHERE clause or statement to subset
the data so observations with missing values for the ID variable are
omitted.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 TS027
CMS 6.09 TS450
CONVEX 6.09 TS027
HP800 6.09 TS027
MVS 6.09 TS450
Solaris 6.09 TS027
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
VMS 6.09 TS450
WIN/NT 6.09 TS027
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
OS/2 6.10 TS019
Windows 6.10 TS019
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
WIN/NT 6.11 TS040
Win95 6.11 TS040
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-CAPABILITY-C519 *** ALERT NOTE *** PRODUCT: QC
PROCEDURE:CAPABILITY
TITLE: Possible incorrect plot in key cell of a comparative histogram
KEYS: KEY CELL COMPARATIVE HISTOGRAM COMPHIST COMPHISTOGRAM DATA MISSING
INCORROUT MIDPOINTS OUTLIERS
If data in the key cell (top left cell) of a comparative histogram are
outside the range of midpoints specified with a MIDPOINTS= option on a
COMPHISTOGRAM statement, then these outlying points will be missing from
the plot in the key cell. This does not happen when cells other than the
key cell contain data beyond specified midpoints. In this situation, the
procedure correctly extends the midpoint list to accommodate the data
ranges in all cells. Note that if outliers occur in the key cell and
non-key cells then the procedure will only extend the axis enough to
accommodate the non-key cell data and points may still be missing from
the key cell.
To circumvent this problem, specify midpoints that span the range of
data values in the key cell, or use the default horizontal axis scaling
by omitting the MIDPOINTS= option.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 TS027
CMS 6.09 TS450
CONVEX 6.09 TS027
HP800 6.09 TS027
MVS 6.09 TS450
Solaris 6.09 TS027
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
VMS 6.09 TS450
WIN/NT 6.09 TS027
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
WIN/NT 6.11 TS040
Win95 6.11 TS040
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MIXED-C520 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: MIXED
TITLE: Incorrect Results From CONTRAST/ESTIMATE With Multiple RANDOM Statements
KEYS: INCORRECT RESULTS CONTRAST ESTIMATE MULTIPLE RANDOM STATEMENTS GROUP
STAT MIXED PROC
Incorrect results can be reported from the CONTRAST or ESTIMATE
statements when multiple RANDOM statements are used. At least one
of the RANDOM statements must use the GROUP= option for this
problem to occur.
The only workaround for this problem is to recode your RANDOM
statements without using the GROUP= option.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
CMS 6.09 TS450
MVS 6.09 TS450
VMS 6.09 TS450
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
OS/2 6.10 TS019
Windows 6.10 TS019
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
WIN/NT 6.11 TS040
Win95 6.11 TS040
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MIXED-C557 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: MIXED
TITLE: Incorrect Predicted Values with Multiple RANDOMs and Non-Nested SUBJECTs
KEYS: INCORRECT PREDICTED VALUES MULTIPLE RANDOM NESTED SUBJECT= OBSERVED
STAT MIXED PROC
Incorrect predicted values from the P or PM options or incorrect
values for the dependent variable can be reported in the PREDICTED
table when multiple RANDOM statements are used with non-nested
SUBJECT= effects.
There is no circumvention for this problem.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 TS027
CMS 6.09 TS450
CONVEX 6.09 TS027
HP800 6.09 TS027
MVS 6.09 TS450
Solaris 6.09 TS027
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
VMS 6.09 TS450
WIN/NT 6.09 TS027
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
OS/2 6.10 TS019
Windows 6.10 TS019
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
WIN/NT 6.11 TS040
Win95 6.11 TS040
AIX/R 6.12 TS005
HP800 6.12 TS005
OS/2 6.12 TS005
Solaris 6.12 TS005
VMS/AXP 6.12 TS005
Windows 6.12 TS005
WIN/NT 6.12 TS005
Win95 6.12 TS005
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-RLIBRARY-C654 *** ALERT NOTE *** PRODUCT: CONNECT
PROCEDURE: RLIBRARY
TITLE: Data set corruption may result with Win95 client and APPC access method
KEYS: SET CORRUPTION CONNECT SNA SERVER MICROSOFT LAN TRANSPORT TCP TCPIP RLS
REMOTE LIBRARY SERVICES W95 WNC CLIENT DATA CORRUPT GARBAGE
Data set corruption may occur on a remote SAS/CONNECT single user
server or on a SAS/SHARE server if the following conditions apply:
1) Using remote SAS/CONNECT single user server or SAS/SHARE server
on a Windows 95 platform with SAS Release 6.11 TS040+.
2) The communications access method being used is APPC.
3) Utilizing Microsoft's SNA Server as the underlying APPC protocol
stack. In particular, Version 2.11+ of the Windows 95 client
software.
4) The Windows 95 client software is configured to use TCP/IP as the
LAN transport.
The only verified circumvention is to configure the Windows 95 client
software to use IPX/SPX LAN transport, instead of TCP/IP. Using
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.09 TS450
MVS 6.09 TS450
VMS 6.09 TS450
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-SERVER-C655 *** ALERT NOTE *** PRODUCT: SHARE
PROCEDURE: SERVER
TITLE: Data set corruption may result with Win95 client and APPC access method
KEYS: SET CORRUPTION CONNECT SNA SERVER MICROSOFT LAN TRANSPORT TCP TCPIP RLS
REMOTE LIBRARY SERVICES W95 WNC CLIENT DATA CORRUPT GARBAGE
Data set corruption may occur on a remote SAS/CONNECT single user
server or on a SAS/SHARE server if the following conditions apply:
1) Using remote SAS/CONNECT single user server or SAS/SHARE server
on a Windows 95 platform with SAS Release 6.11 TS040+.
2) The communications access method being used is APPC.
3) Utilizing Microsoft's SNA Server as the underlying APPC protocol
stack. In particular, Version 2.11+ of the Windows 95 client
software.
4) The Windows 95 client software is configured to use TCP/IP as the
LAN transport.
The only verified circumvention is to configure the Windows 95 client
software to use IPX/SPX LAN transport, instead of TCP/IP. Using
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.09 TS450
MVS 6.09 TS450
VMS 6.09 TS450
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS040
WIN/NT 6.11 TS040
Win95 6.11 TS040
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MIXED-C658 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: MIXED
TITLE: Variance Components Incorrect for FA(0) With RANDOM Statement
KEYS: VARIANCE COMPONENTS INCORRECT FA(0) RANDOM STATEMENT STAT PROC MIXED
The variance components reported by the RANDOM statement with TYPE=
FA(0) are incorrect. They should by multiplied by the SQRT(RESIDUAL).
If the NOPROFILE option has been used or if a REPEATED statement is
also used, then the variance components will be correct. To circumvent
this problem, use the NOPROFILE option on the PROC MIXED statement.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 TS027
CONVEX 6.09 TS027
HP800 6.09 TS027
Solaris 6.09 TS027
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
WIN/NT 6.09 TS027
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
OS/2 6.10 TS019
Windows 6.10 TS019
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MIXED-C659 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: MIXED
TITLE: Variance Component for CSH Parameter Can Be Incorrect
KEYS: VARIANCE COMPONENT CSH RANDOM STAT MIXED INCORRECT PROC
The variance component reported for the CSH parameter can be incorrect
when using TYPE=CSH on the RANDOM statement. The estimate should be
divided by the residual variance estimate. This error will not
occur if the NOPROFILE option is used or if a REPEATED statement is
also used. To circumvent this problem, use the NOPROFILE option on
the PROC MIXED statement.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404 6.09
MVS 6.08 TS404 6.09
VMS 6.08 TS404 6.09
VSE 6.08 TS404 6.09
AIX/R 6.09 TS027 6.11
CONVEX 6.09 TS027 6.11
HP800 6.09 TS027 6.11
Solaris 6.09 TS027 6.11
ULTRIX 6.09 TS027 6.11
VMS/AXP 6.09 TS027 6.12
WIN/NT 6.09 TS027 6.11
IABI 6.10 TS018 6.11
MIPS ABI 6.10 TS018 6.11
Macintosh 6.10 TS037 6.11
OSF/AXP 6.10 TS018 6.11
OS/2 6.10 TS019 6.11
Windows 6.10 TS019 6.11
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MIXED-C661 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: MIXED
TITLE: RATIOS Column in Models with FA(0) Structures Incorrect
KEYS: RATIOS COLUMN MODELS FA(0) VARIANCE COMPONENTS INCORRECT RANDOM MIXED
PROC STAT
The RATIOS column in models with a RANDOM statement and TYPE=FA(0)
will be incorrect. The square root of the residual variance is
used to calculate the ratios, rather than the residual variance itself.
To circumvent this problem, use the NOPROFILE option on the PROC MIXED
statement.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
AIX/R 6.09 TS027
CMS 6.09 TS450
CONVEX 6.09 TS027
HP800 6.09 TS027
MVS 6.09 TS450
Solaris 6.09 TS027
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
VMS 6.09 TS450
WIN/NT 6.09 TS027
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
OS/2 6.10 TS019
Windows 6.10 TS019
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
WIN/NT 6.11 TS040
Win95 6.11 TS040
AIX/R 6.12 TS005
HP800 6.12 TS005
OS/2 6.12 TS005
Solaris 6.12 TS005
VMS/AXP 6.12 TS005
Windows 6.12 TS005
WIN/NT 6.12 TS005
Win95 6.12 TS005
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-SYS.PROC-C682 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: SYS.PROC
TITLE: Incorrect Derivative When Sum Statement Used in PROC MODEL or PROC NLP
KEYS: INCORRECT DERIVATIVE SUM STMNT MODEL NLP NLIN
The compiler used in the MODEL and NLP procedures will compute
incorrect analytic derivatives when a sum statement is used in the
model specification. The sum statement is often used in a summation DO
loop. For example, the following code would return an incorrect
derivative for y:
y=0;
do i=1 to 3;
y + x + a;
end;
To circumvent the problem, replace the sum statement with an assignment
statement to define the summation. The above example would be modified
as
do i=1 to 3;
y=y + x + a;
end;
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
WIN/NT 6.11 TS040
Win95 6.11 TS040
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-AF-C942 *** ALERT NOTE *** PRODUCT: AF
PROCEDURE: AF
TITLE: Values set in INIT of the model SCL may not be saved
KEYS: DATA FORM TABLE DATAFORM DATATABLE FRAME INIT MODEL SCL VALUE NOT SAVED
RETAINED SET
Values set in the INIT section of the Data Form or Data Table's model
SCL entry may not be saved. This occurs if multiple columns are
referenced anywhere in the model SCL and columns are updated in the
INIT section. In this case, INIT will only update a column if, in
the order of the variables in the data set, this column appears
first in relation to the variable position of the other columns
referenced in the model SCL. This is only a problem in INIT.
For example, in the following code fragment for model SCL using the
SASUSER.CLASS data set, neither the columns AGE nor HEIGHT will be
updated since the column NAME was accessed in the model SCL and it comes
before AGE and HEIGHT in the order of the variables in the data set.
INIT:
age=55;
height=55;
return;
MAIN:
put name=;
return;
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
AIX/R 6.12 TS005
HP800 6.12 TS005
OS/2 6.12 TS005
Solaris 6.12 TS005
VMS/AXP 6.12 TS005
Windows 6.12 TS005
WIN/NT 6.12 TS005
Win95 6.12 TS005
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-MODEL-C938 *** ALERT NOTE *** PRODUCT: ETS
PROCEDURE: MODEL
TITLE: GMM incorrectly handles model with negative sign on intercept term
KEYS: GMM INCORRECTLY HANDLES MODEL NEGATIVE SIGN INTERCEPT TERM INCORROUT
INCORRSTAT CONVERGENCE UNABLE TO CONVERGE FIT
If you use GMM to estimate the parameters of a model in which a hard-
coded negative sign is associated with the intercept term, such as:
y = -a + b*x;
then PROC MODEL may either return incorrect results or have difficulty
converging to a solution.
To circumvent the problem, reparameterize the model specification so
the intercept term does not have a negative sign associated with it.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 TS027
CMS 6.09 TS450
CONVEX 6.09 TS027
HP800 6.09 TS027
MVS 6.09 TS450
Solaris 6.09 TS027
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
VMS 6.09 TS450
WIN/NT 6.09 TS027
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
OS/2 6.10 TS019
Windows 6.10 TS019
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
WIN/NT 6.11 TS040
Win95 6.11 TS040
AIX/R 6.12 TS005
HP800 6.12 TS005
OS/2 6.12 TS005
Solaris 6.12 TS005
VMS/AXP 6.12 TS005
Windows 6.12 TS005
WIN/NT 6.12 TS005
Win95 6.12 TS005
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-GLM-C842 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: GLM
TITLE: DUNCAN option on MEANS statement gives wrong critical values
KEYS: DUNCAN OPTION MEANS STATEMENT GIVES WRONG CRITICAL VALUES
INCORROUT INCORRSTAT GLM ANOVA STAT
The critical values for the DUNCAN's test should be monotone non-
decreasing in the number of means. Occasionally they are not, if the
CLASS variable has "many" LEVELS. This is a machine dependent problem,
but no machine should have a problem with fewer than 30 means.
To circumvent the problem one may try other multiple comparison tests,
such as TUKEY, LSD.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
Apollo 6.07 TS101
AIX/R 6.07 TS101
AViiON 6.07 TS101
BULL 6.07 TS201
CMS 6.07 TS301
HP300 6.07 TS101
HP800 6.07 TS201
MIPS 6.07 TS101
MVS 6.07 TS301
NeXT 6.07 TS201
SGI 6.07 TS201
Siemens 6.07 TS201
Sun-3 6.07 TS201
Sun-4 6.07 TS201
ULTRIX 6.07 TS101
VMS 6.07 TS301
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 TS027
CMS 6.09 TS450
CONVEX 6.09 TS027
HP800 6.09 TS027
MVS 6.09 TS450
Solaris 6.09 TS027
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
VMS 6.09 TS450
WIN/NT 6.09 TS027
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
OS/2 6.10 TS019
Windows 6.10 TS019
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
WIN/NT 6.11 TS040
Win95 6.11 TS040
AIX/R 6.12 TS005
HP800 6.12 TS005
OS/2 6.12 TS005
Solaris 6.12 TS005
VMS/AXP 6.12 TS005
Windows 6.12 TS005
WIN/NT 6.12 TS005
Win95 6.12 TS005
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-REG-C941 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: REG
TITLE: Incorrect Cp and BIC values may be given on plot
KEYS: INCORRECT CP BIC PLOT INCORROUT INCORRSTAT
If you request the CP and/or BIC options on the PLOT statement and
you are using SELECTION=MINR or SELECTION=MAXR, the values that are
are placed on the plot are incorrect. The values that are reported
elsewhere in the output are correct.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
AIX/R 6.12 TS005
HP800 6.12 TS005
OS/2 6.12 TS005
Solaris 6.12 TS005
VMS/AXP 6.12 TS005
Windows 6.12 TS005
WIN/NT 6.12 TS005
Win95 6.12 TS005
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-GLM-C889 *** ALERT NOTE *** PRODUCT: STAT
PROCEDURE: GLM
TITLE: PROC GLM Degrees of Freedom can be wrong if SS2 SS3 SS4 is specified
KEYS: PROC GLM DEGREES FREEDOM WRONG SS2 SS3 SS4 SPECIFIED STAT INCORROUT
INCORRSTAT CHANGE DF MODEL
Prior to SAS Release 6.12, if one specifies more than SS2, SS3, SS4
on the MODEL statement in PROC GLM the degrees of freedom (DF) for all
of them are the same and equal to the DF for the highest SS computed.
For example, if MODEL Y=A B A*B/SS1 SS2 SS3 SS4. PROC GLM will
report the SS4 DF for the SS2 and SS3 DF.
This is a problem when the true DF for the different SS options are not
equal to the DF for the highest SS computed. To see if one has
encountered this problem it will be necessary to run a separate GLM
for each type of SS requested on the MODEL statement and compare
the DF to those reported when more than one SS is specified
on the MODEL statement.
To circumvent the problem, specify a separate GLM for each type
of SS requested.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.07 TS301
MVS 6.07 TS301
VMS 6.07 TS301
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
AIX/R 6.09 TS027
CMS 6.09 TS450
CONVEX 6.09 TS027
HP800 6.09 TS027
MVS 6.09 TS450
Solaris 6.09 TS027
ULTRIX 6.09 TS027
VMS/AXP 6.09 TS027
VMS 6.09 TS450
WIN/NT 6.09 TS027
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
OS/2 6.10 TS019
Windows 6.10 TS019
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
WIN/NT 6.11 TS040
Win95 6.11 TS040
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-REPORT-C829 *** ALERT NOTE *** PRODUCT: BASE
PROCEDURE: REPORT
TITLE: PROC REPORT may display incorrect computed values when scrolling
KEYS: PROC REPORT DISPLAYS WRONG STATS PAGING FORWARD BACK PAGE STATISTICS
NUMBERS INCORROUT
If you scroll through a multi-page report in the PROC REPORT window
using the PAGE and PAGEBACK commands repeatedly, the values of some
computed columns may be recalculated incorrectly.
It is possible to correct the values by returning to the first page of
the report, and then using the PAGE command to scroll forward again.
The problem only occurs when you scroll forward and back several times.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
CMS 6.08 TS404
MVS 6.08 TS404
OS/2 6.08 TS404
VMS 6.08 TS404
VSE 6.08 TS404
Windows 6.08 TS404
CMS 6.09 TS450
MVS 6.09 TS450
VMS 6.09 TS450
IABI 6.10 TS018
MIPS ABI 6.10 TS018
Macintosh 6.10 TS037
OSF/AXP 6.10 TS018
OS/2 6.10 TS019
Windows 6.10 TS019
AIX/R 6.11 TS020
HP800 6.11 TS020
IABI 6.11 TS040
MIPS ABI 6.11 TS040
OSF/AXP 6.11 TS040
OS/2 6.11 TS020
Solaris 6.11 TS020
Windows 6.11 TS020
WIN/NT 6.11 TS040
Win95 6.11 TS040
AIX/R 6.12 TS005
HP800 6.12 TS005
OS/2 6.12 TS005
Solaris 6.12 TS005
VMS/AXP 6.12 TS005
Windows 6.12 TS005
WIN/NT 6.12 TS005
Win95 6.12 TS005
Return to top
>>>>>>>> SAS Notes -- Copyright (c) 1991 - 1996 by SAS Institute Inc. <<<<<<<<
V6-EIS-C906 *** ALERT NOTE *** PRODUCT: EIS
PROCEDURE: EIS
TITLE: Graphical Variance report may misrepresent the data
KEYS: EIS GRAPHICAL VARIANCE REPORT GVAR OBJECT DATA BARS INCORRECT
MISREPRESENTED INVALID VALUES
The Graphical Variance Report object in SAS/EIS Software may display
the top subgroups of the bar in the incorrect color. This may occur when
the chart contains a mixture of bars representing both GOOD and BAD
results. For example, all bars may appear to represent GOOD results
when some bars actually should represent BAD results and vice-versa.
+------REPORTED------+ +---FIXED---+
SYSTEM RELEASE LEVEL RELEASE LEVEL
AIX/R 6.12 TS020
HP800 6.12 TS020
OS/2 6.12 TS020
Solaris 6.12 TS020
VMS/AXP 6.12 TS020
Windows 6.12 TS020
WIN/NT 6.12 TS020
Win95 6.12 TS020
Return to top