ASReml v4.1: Release Notes

ASReml 4 Version History

ASReml 4.1 is the current released version
The documentation folder contains the UpdateR4.pdf file that identifies what's new in Release 4.1
In addition there are two ASReml 4.1 User Guide document files available :- UserGuideStructural.pdf and UserGuideFunctional.pdf respectively

v4.1mx (30th September 2018)

!TAU is allowed as an alias for !Beta
A bug in parsing xfa1(leg(bbb,1)).Subject has been fixed.
A bug in UPCASE (overflow linelength) has been fixed.
A bug in pol/leg functions with missing X variables has been fixed.
A notice has been added if TSV changes constraints.
A bug when there are tab separated tokens in .pin file has been fixed.
A bug in processing at(year,2).jlast when !MVREMOVE has been fixed.
The order for reordering Cinverse has been changed.

v4.1mw (31st May 2018)

The use of # in is now allowed in pedigree entry names.
Improved .tsv error messages.
A problem in parsing !SUBSET has been fixed.
A problem when grm(Mcln) has levels coded 0 has been fixed.
A problem when GIV has many more levels than base !I factor has been fixed.
A problem that giv referenced the wrong variable when factor defined after a GROUP factor has been fixed.
Fix to failure to honour second trait weights when estimating missing values.
A problem causing ASReml to crash while reporting the variance components when using maximum workspace has been fixed.
New unofficial qualifiers (for investigation) !Beta and !ARlimit.

v4.1mv (28th November 2017)

The vec() model function, which applies a set of variables (!G g) in parallel to a set of g traits, was not working when the variables were factors (!G g l) coding for l class; this has now been fixed.
The number of lines that report singularities in a G matrix has been reduced, reporting them as groups rather than each singularity separately.
A bug has been fixed, which meant ASReml failed to read a GRM matrix if it was stored full.
A labelling problem has been fixed, to label components properly when out of order.
Some warning notices have been modified to make their meaning clearer.
A bug, which allowed a correlation to update to zero, has been fixed; a correlation parameter of zero is not allowed.
Contents of the .sco file in VCM cases have been modified to report SCORE and constrained GAMMA from the final iteration.
Processing of 3-way models, involving reduced rank factor analytic (F) and a GRM matrix (G), has been revised. The processing rules require that such models are specified with these terms respectively first and last; i.e. F.I.G or G.I.F. Messages reporting failure of this requirement have been updated.
G.A.F models can now be fitted. An updated error message is given, if a VCM matrix is not positive definite.
A bug has been fixed in utilizing ASSIGN strings when one string name is a shortened form of another (e.g. using t and tot).
There has been a general revision in the processing of VCM constraints on variance parameters, to improve handling of constraints applied to different types of parameters.
An integer overflow bug, that occured in calculating a Deviance table when a large amount of memory was utilized, has been fixed.
Adding of 'hidden' model terms, when !WRITEDESIGN is specified, is now avoided, since prediction is not performed in such jobs.

v4.1mr (31st May 2017)

The notice associated with use of the !GOFFSET (pedigree groups) qualifier has changed. If this qualifier is used, it is now recommended you try several values for the parameter value (say 0.1, 1 and 10), and choose the value which gives the highest likelihood. Use of a small value when genetic group effects are small can bias the genetic variance down.
Previously the syntax sat(ENV).units failed when ENV had more than 999 levels. Now the limit is 9999 levels; but ASReml can not fit that many variance components.
ASReml now detects when a model term is actually larger than anticipated.
Previously ide(Sire.CG) crashed ASReml. ide() was not intended to be used this way as the number of levels may not be set correctly; the term should be specified as ide(Sire).CG when a NRM is associated with Sire.
Several bugs have been fixed in relation to the !VCM qualifier. This procedure can link parameters of different types which can confuse the parameter update rules. The VCM matrix is printed to the .asl file in !DEBUG mode to facilitate checking.
Other bug fixes relate to spelling, mishandling of missing values in some situations, parsing !DO !ENDDO transformations and some graphical output when processing multiple parts.

v4.1mn (10th January 2017)

bug in !RANGE qualifier so it works with data -10 : 10.
Fix inconsistency in AIVAIG arg list affecting XLINK.
Change SLIST(100), CLIST(100) to allocatable.
Bug labelling !A factor when more than 90000 !A levels.
Bug in Stratum Var when variances large.
Default plot format changed to EPS.
Comment if var(Y) extreme.
Allow xfa x ar1 x ar1.
Bug setting up residual sat(Site).id(units).diag(2).
Failed to honour !S2==1 in multisection multivariate STRUCTURAL w/o Start values
Failed to reset REREAD in MS MV STRUCTURAL w/o SV.

v4.1mj (1st July 2016)

!SUM requests additional distributional information and the information given depends on whether the variable is a factor or a variate. But a crash occurred when a variable is declared as a factor and a variate. The fix is to treat it just as a factor (the primary definition).
Fix a bug causing a CRASH when the !SUMMARY qualifier is specified to request distributional information on variables and a factor (say Row) is also fitted as a variate (lin(Row)). The variable is treated as a factor

v4.1mh (10th June 2016)

Better handling of correlation parameters approaching 1.0.
Fix calc of trace part of SCORE and of WV for DIAG term in RR.I.DIAG.
Honour !GF in .as over changed value in .rsv.
Fix failure in VPREDICT with long modelterm name.
Fix STACK overflow and problem forming LogDet for large GIV matrices.
Fix failure in VPREDICT with negative correlation in CORUH.
Fix failure to predict with SUBSET Trait.

v4.1mg (28th April 2016)

In the oats example, Nitrogen !A !L 0_cwt 0.2_cwt 0.4_cwt 0.6_cwt thus now works.
The special factor Trait can now be included in the PRESENT list. So PREDICT GENOTYPE !PRESENT Trait Location now works, when Location is a fixed factor nested in Trait.
diag(Loc).Variety now works when Loc has only 1 level.
An error has been corrected in a message: 'some time take' is now 'take some time'.
The default PIN file name has been corrected, so that it is now e.g. basename.pin rather than basename.as.pin.

v4.1mf (1st April 2016)

at(Tcode,1).ide(animal !INIT -.1) previously failed because ide was not formally recognised as a function wrapper function.
us(Tcode).nrm(Animal) previously failed when Tcode was a variate; a factor is expected.
and(a.b)us(Trait).Genotype previously failed because a SPACE (or PLUS or COMMA) is required after and(a.b). The and() model function may not interact with another term.
ASReml internally maintains 10 random sequences. !SEED now initializes all sequences rather than just the first.
The XML tag ZRatio was misspelt as Zratio in one place.
Previously, the number of elements in the MBF file was limited to 10,000. This limit has been removed so that a user can process a large marker file through this route. However, the GRR machinery will generally be much more efficient.
The format of the temporary file used when the user specified the factor labels has been modified so that up to 99,999,999 labels may be supplied (previous limit was 999,999).
When the pedigree was read from a data file with very long lines (more than 1000 characters), it was misread. This issue is now fixed.
A new variance wrapper function, rr() has been added to facilitate specification of XFA models with all zero PSIs. rrk(Site) is equivalent to xfak(Site s * 0 < Initial Loadings > ), where s is the number of levels of the factor Site. The point of this change is that often ASReml will take much longer to fit the term xfak(Site).Genotype than to fit the equivalent diag(Site).Genotype rrk(Site).Genotype.

v4.1mc (4th November 2015)

A problem reading the data that occured, when data lines were very long but only a few data fields were being read, has been resolved.
There was a problem initializing the list of factor level labels, when more than 1 million were defined. The limit has been increased to 100 million.
There was a problem resulting in an ASReml crash, when plotting residuals versus fitted values from a GAMMA/INVERSE GLMM when missing values were fitted.
There was a problem recognising a VCM statement in the .as file, when the RESIDUAL statement was omitted.
The fault message, when ASReml failed to form the mixed model equations because of insufficient work space, was misleading. The messages associated with faults 333 and 315 have been revised.
!EMFLAG now defaults to parameter value 5 (local EM steps).
A job fitting a model term xfa1(Section).grm(Entry), with 8 sections and 600 Entries where grm(Entry) is dense, seemed to take longer than expected. It turned out that the equivalent model xfa1(Section 8*0 8*0.1).grm(Entry) + diag(Section).grm(Entry) ran more than 10 times faster. Setting the specific variances to zero is now equivalently done with the rr() variance function, so the equivalent model may now be written as rr1(Section).grm(Entry) + diag(Section).grm(Entry)
Having rectified a problem of large compute times for PEV calculation for marker effects in GBLUP analysis, the PEF values are now reported by default. Use !NOPEV on the .grr line to suppress them.

v4.1lr (10th March 2015)

Cumulative enhancements from internal version 4.1lc - 4.1lr
a4data Allow MODIFY to work on nrm(Animal .01 !GU)
!CONTINUE and !LAST problems following qualifiers
Allow multiple increases in workspace allocation
Change to Long Integer addressing for XFA setup
Fix crash in VWVG with XFA large memory
Fix failure to handle XLT internal function in loop when large memory
Convert Address from short to long (W20 crash Weights)
Test VPREDICT has 3rd field; so no CRASH
Fix bug that tabulate not honoured when it follows RESIDUAL
Fix error; mean lines reversed in !X !Y plots
Add LinePrinter version to .res file
Stop stripping wrapper from simple terms
SITE*6 now becomes SITE*8
If name given for residual, use it e.g. units;Residual
Allow lit;us[1:3] in VPREDICT
Fix bug reading xfa(fac,2).lit
Fix bug parsing XFA structural
Insert THRLEVELS in ARGlist c.f. lg/knwd (needed for Cheese example
Catch bug when logical table has more than 4225000000 cells
Delete erroneous position message for idv(A) variance structure
Recognise xxx.grr Factor
Change to us(T).xxx;us(T) naming of components.
Change !CENTRE options
Bug recognising GenVar when GenVar created Sire*4
Update syntax NOKEY => NOID etc.
!CENTRE activated
Fix bug in DenDF calculation for ANTE1 in grass.as
Extend .xml to .pvs output
Extend level labels for 3-way interaction when !SLNFORM is set
Enable !SMODE 32
convert mkr/mat to .grr
convert snp() to grr()
GRR file initializes Factor class names if GIDname matches an !A variable name
Allow VPREDICT to use labels
Allow !FINAL on data line
Review LofDet calc of GIV
Fix bug reporting variance parameters in .res
!AIF replaces !DIAG
!XML to request XML output
Grid infill now works for multiple sections
Add !XML Command Line qualifier
Fix bug in restoring label IDV_V
Fix bug to eport final itern seq when simple model
Add !FINAL
Hold repeated level lists more efficiently
Fix bug reading in DENSE GIV of order 3189
Extend mvs example to !ASUV
Correct fix for threshold model examples (Cheese)
Update to .mat input option
Fix bug parsing !LAST Identifier.us(Trait)
Fix problem labelling correlations of 0/1 variates
Fix weighted multivariate analysis failure
Fix bug parsing -pSMTTWNP_30 smt (2nd P confused it)
Fix wrongly applied US to spl(EWEAGE) in thr 4 model
Fix LogL for weighted multivariate analysis
Fix PREDICT EWE !PRESENT BFLOCK under threshold model
Modify sln header layout
Fix bug in writing bss.vvp
Adjust workspace check near line 2653
Refine NGAMMA count in bss.as job (SING)
Modify SITE from A6 to A8 so better labeling in plots
Modif SITE from A6 to A8
Change bgrm to sgrm to match documentation

v4.0lb Beta (28th July 2014)

Cumulative enhancements from internal version 4.0ks - 4.0lb
Get start values xfa(A,1) ==> us(A)
Fixed bug parsing !SUBSET Tr123 Trait 1 2 3 0 0
Fixed bug in forming PEV of Selection Index
Fixed crash when invalid model term specified
Fixed SE reported when Singular component
Adjust switching from 1 to multiple factor update
Exclude factors from XFA eigen analysis
Fixed confusion parsing xfa(F,k)
Insert default initial values for XFA structures
Fixed confusion when user specifies ?sv file to read
Fixed bug: not opening .rsv file
Extended VPREDICT syntax
Better scaling in .vvp
Fixed bug parsing xfa2(Trait)
Fixed bug in reading named .tsv file
Adjustment so GDENSE works with singular GRM
Fixed to avoid crash when RES/PV is NaN
Inserted heading into .sln file
Fixed to avoid crash on diagnostic print when few (<40) equations
Adjusted format of diagnostic print
Fixed to print warning when $string undefined
Fixed new bug in !ASSIGN testing
Fixed new bug in testing for partial matches
Fix to disallow special characters in new factor names
Fixed format 2470, change condition
Prevented MKR BLUP overwriting KCV
Test factor labels of SUBSET better
Test Assign labels better
Increased space for GRM matrix
Report correct name of .rsv file used
Changed logical units to avoid conflict when CYCLE HOLD
Reduce XFAIF when MAXIT=1 (assume near solution)
PEV of marker blups (mkr/grm): now works
Extended VPREDICT to S X C D functions
Fixed .vvp for XFA on gamma scale
Fixed bug parsing labels on marker file when not many of them
Replaced SMIGE with SMISG so Singular GRM allowed
Added !PSD qualifier to declare number of singularities in GRM
Added !HOLD qualifier to !CYCLE so it only reads data and mkr file first cycle
Made !ASUV default for BINNOR model
Made !SPECIALCHAR parsing more robust
TPREDICT hyb !AVE xfa1(site) !ONLY xfa1(site).hyb
Allowed negative pivot in Dense equations (e.g. GRM)
Report more characteristics of GRM
Enabled reading labels from .mkr files
Improved histogram for mkr file report
Fixed bug when long line splits as NA*
Increase size of CCOL/CVAL work vectors to avoid overflow with and()
Fixed failure to parse and(Lhyb.xfa1(Loc),-0.2)
Added notice for MODIFY sat(Trial)... case
Revamped TSV/RSV/CONTINUE
Fixed bug reading .rsv when two xfa model terms
!FOLDER can now appear before variable definitions
Class names file may be in FOLDER
Added !SUBSECTION diagnostic warning if too many sections
Updated Data Sort parameters; modified report of 'size' of units
Fixed bug when and() term IGNORED
Added warning if !SUBSECTION result is suspicious
Added code to fix class names appearing as NA
Let !SECTION !ROW !COL inform record sorting if residual model does not
Fixed bug labelling XFA factors
Inserted code for if class levels NA
Fixed bug Recognise !A !L ../filename
Fixed bug in '!CONTINUE !OUTFOLDER'
Check CONTRAST basis variable: warning
Implemented new ASReml license
Fixed bug: failed to sort xfa.ped equations
Increased default space for sorting under !ROWFAC !COLFAC
Increased HELPFL to 200 characters
Fixed check of at() str too long; allow 190 rather than 90
Clear extra characters when expanded length > 1000
Changed LogDet calc message to Notice
In !L file, allowed "Chap 5" ==> 'Chap_5'
More robust LDET of sparse GIV
Fixed bug opening named Scratch file
Fixed bug when Scratch file not large enough
Fixed misreading long lines (character dropped)

v4.0kr Beta (26 Mar 2014)

Fixed format bug; truncated counts when NREC<99999
Fixed bug reporting XFA matrix in .asr when some PSI zero.
Fixed LongInt addressing Bug reporting spl trend with Huge workspace
Fixed bug producing Predict PLOT with !PARALEL qualifier

v4.0kq Beta (16th March 2014)

Tighten model parsing checks and( , 0.2)
Allow residual units.mthr(Trait)
remove IF rgmstr==4 so .msv has same structure as .tsv
moved 'CALL StructureCheck' so ARCOL.as 18 now runs
modified StructureCheck to insert missing V parameter in G structures
Bug reporting pedigree conflict fixed, !UPPER added
!UPPERCASE added
Implement !GINDEX as TPREDICT
Added message when US not PD
Cancelled trying to reuse previous equation order: failing

v4.0kp Beta (5th March 2014)

Cumulative enhancements from internal version 4.0kn - 4.0ko
Fixed Bug parsing long data lines
SLASHSLASH restricted to .as file
Parent/Non-Parent labeling in aif file
NPD US message updated
GDENSE default
failed on at(parity).units.us(Trait)
Stop if more geno than markers
Add processing of decimal markers
leave log() and sqrt() wrappers
Fixed Bug getting DET of GIV with Groups

v4.0km Beta (9th January 2014)

Cumulative enhancements from internal version 4.0kk - 4.0kl
Fixed Bug parsing !FOR $LIST !DO command
Fixed Bug reading binary MBF file
Fix Bug creating design for row and(col)
!SIGMAP to allow the Sigma parameterization
No longer strip v from idv(units) under gamma parameterization
Fix Bug processing nrm(animal !INIT .25); fill in STRUC 14,16
Fix Bug processing idv(units.Trait) (!ASUV)
Added KEEP (ASUV flag) to SETSTR arg list
Do not create MODIFY lines for '-idv' terms
Set !SIGMAP for xxh() Rstructure models
Fix BUG in RCOV address (orange spl() error)
Fix fault Pedigree sort failure with Genetic Groups

v4.0kj Beta (17th January 2014)

Fix so !SUBSET Trait 3 4 5 0 0 has factor size 3
Fix misspelling of constraint
Comment out debugging statement
Fix Bug resetting RSVFILE with !WORK
Fix Bug parsing !CONTINUE rsvfile
Extend !AIPENALTY
small changes to bsfx, mssyfa1, xfab
Fix Bug switching back to all factors when 2 XFA terms!
Moved !DOPART $1 to top command line!
Increased Rstring and Selectstrg to *24
Inserted factor namr in title of VG plot

v4.0ki Beta (7th January 2014)

Cumulative enhancements from internal version 4.0kb - 4.0kh
Give heading for ar1v() term in component table
Fixed Bug parsing at(Age):ar1(Row):ar1(Col):us(Trait)
Fixed bug arising from new VC table labelling.
Needed to convert I4 to I8 integers KPLUSW, MEMORY
Long LXBASE+AIOPT in airhat call
Modified format 2133 and rearranged WRITEs using it
Fixed bug reporting conflicting pedigree (labelling)
Fixed bug parsing when reading pedigree
Modified variance components table at request of Robin extensive output changes.
Bug: inserted MOD() for case of -ve mdel term
Extended functionality of VCM for Christ Brien
Extended VCM statement to allow list of parameters in place [VCB]
reordered so license check before prompts
Do not set space for data sorting if not required
Do not zero sln(EQN,2) as it sometimes crashes asreml-r
Change base 16000 to 17500 so can have more than 1000 loadings
Allow I9 for counts if MROW > 99999 (else I6)
!PRECISION 1 2 3 === 6 5 4
Assume random for nrm, grm and vs structures
!SAVE equivalent/alternative to !GIV
!L filename !SKIP equivalent to !LSKIP
Accommodate spurious leading BLANKS
Allow blank lines in files being merged
Notify if no Trait with !MULTINOMIAL
Changed comment
Changed LU=8 to LU=13 so TABULATE works!
Revampted Variance component table.
Bug adjusting arh(row)
Adjusted MP absorption so return correct Fincr
Fixed bug parsing ped col labels ANIMAL,SIRE,DAM
Improved error messages
Inserted links for !SUBSECTION in new syntax
Avoided bad memory reference
Fix format bug
Allow '*' on next line parsing old format Structure lines
Bug parsing !CONT 3
Bug reporting restart from .msv
Bug parsing str(row
Added RSVFILE
Allowed filename arg to !CONTINUE for .rsv file
Added warnings in two places where // notation fails.
Fix labelling of Matern parameters - Matern.as

v4.0ka Beta (08th October 2013)

Cumulative enhancements from internal version 4.0jy - 4.0jz
Extended code to permit residual diag().ar1().ar1()
Inserted 'Notice:' in diagnostic message
Fixed bug in Peppers job by moving VSOLN(EQN)=XD
Modified labelling of pedigree counts table.
Fixed bug wrt flevel(quadGF) when quadGF is zero
Treat MKR data outside -2,2 as missing.
Faster parsing of marker file values 0,1,2
Fixed bug in default skipping of first field of .mkr file
Fixed bug making file names for !out option with !RENAME cycling
Fixed bug when Insert Exit path when !A !L labels file not present.
Read error diagnostic print added
Added links for !EXCLUDE and !KCV
Added actions for !EXCLUDE
Added reading !EXCLUDE and !KCV qualifiers
Added KCV functionality
Added KCV and ExCLUDE variables
Fixed bug due to unset variable KS.

v4.0jx Beta (08th September 2013)

Cumulative enhancements from internal version 4.0ju - 4.0jw
Fixed bug in LogL calc for NSECT>1 when NSPAT>1
[Modifies reported LogL in dhglm, az06, binnor] [Forum query re spatial LogL]
Fixed bug with string overflow when parsing CALC statements (ININTR)
Get GLMM weights in correct order when data records sorted
Suppress erroneous messages parsing pedigree with !MGS option
Take PARENT Labels from 1st line of ped file where provided.
Move .msv file to logical unit 18.
Adjust XFA update cycling parameters.
Fixed bug when GLMM and !{ !} fixed.
Revise XFAIF to incorporate !AIPENALTY
Adjusted limits for pi in !FBB
Tidy up some labeling of VC table
This affects most .asr files as the 'Variance' label has been changed to 'Residual'
Adjust model label strings
!MAXIT -1 aborts run after parsing job and writing .tsv
Added !TSV and !MSV as alternatives to !CONTINUE 2/3
RPscale defaults to 1.0 if supplied as zero
Supply better start values for US residual in !ASUV case
Fixed bug in processing singular VCM matrix
Changed reporting and reading of unset RP fields
Changed so !MAXIT 0 just parses the .as file, writes .tsv
Added !FBB pi to explore Bayes B like models
Changed output heading to Variance Component table affects all jobs
Changed output labeling of components affects SpatialModel jobs
Modified XFA update rules changes Factor Analytic jobs
Modified contents of .tsv affects all jobs
New .msv file for all jobs

v4.0jt Beta (8th July 2013)

fixed license problem in ASReml-W

v4.0jt Beta (2nd July 2013)

fixed bug reporting Boundary parameter as S
fixed bug reading !SPATIAL
fixed bug failing to properly initialize some parameters
fixed Test for label 1E-8
fixed TXTFORM wrong for multiple tabulates
fixed correlation close to boundary
fixed bug plotting residuals from AR.AR.US model
fixed bug parsing residual AR1(R).ar1(C).us(Tr)
fixed bug making predict design for dev(Year).us(Trait)

v4.0js Beta (14th June 2013)

Cumulative enhancements from internal version 4.0jf - 4.0jr
Fixed problem when bivariate weighted used !TOT !WT
Fixed problem of at(Trait,2).X anulling Y for other traits when X missing
Added warning when CSV file has less fields than expected
Allowed qualifiers on idv() terms
Allow !PART list to extend beyond character 80
Honour !VARSCALE for 'residual units' and allow on model line
Fix bug parsing at(evt,'SFM_138)
Fix bug fixing ORDER parsing at(evt,'SFM_138)
Extend !CONTINUE 2 to read .tsv file
Get better default initial values for XFA models
Fix bug in format of MODIFY error message
Fix bug putting AINV structure on geno in loc.geno
Fix bug sorting data for spatial residual multi sections
!PRECISION ; adjust singularity detection threshold in GRM inversion
Change default init var under variance scale
Added !GDENSE qualifier to process GRM() densly
Add !OUTFOLDER qualifier to top command line
Fix bugs processing cholkc() wrapper function
Fixed bugs parsing spherical and circular models
Changed scatter plot to grid base when > 10000 points
Added Mean Axes to XYG scatter plot
Honour !GU after at(G,1).parc
Fix bug in reading G structure terms from new format .rsv file.
Add message when explicit V in R structure removed
Add !CENTRE qualifier, revise weight expression

v4.0je Beta (29th May 2013)

Fixed some license problems.

v4.0jd Beta (4th March 2013)

Beta release of ASReml 4.