ASReml v4.3: Release Notes

ASReml 4 Version History

• ASReml 4.3 is the current released version
• The documentation folder contains the User Guide and what's new documentation

Version History for Release 4.2

v4.2ni (April 2023)

• Now allow row/column labels in dense GRM matrix input
• An update to the licensing for improved error handling
• Added reporting of the coefficient of determination
• Improved speed for processing large predict jobs
• Fix to predict where large data problems could produce a memory error
• Fix to an issue when !XLINK is used with !GROUPS
• Fix to problem processing a PIN file when variance components are constrained to zero

v4.2nh (October 2022)

• New !Hinv qualifier forms an additional special G inverse known as a H inverse
• Added option to calculate LogDet of SPARSE GIV matrix and report negative pivots
• Revised equation ordering to speed up BLUP analyses
• Updates to xml file generation
• Fix to predict where aliased predictions were calculated incorrectly when some random terms were constrained to zero

v4.2ng (February 2022)

• Improvements to fitting of large models (>45K GRM)
• Fix to generation of XML files
• Further improvements to speed of equation ordering
• Improvements to !CINV

v4.2nf (August 2021)

• Improvements to speed of equation ordering
• Improvements to speed of Cinverse formation

v4.2ne (June 2021)

• Improvements to speed up predict process
• Fix to problem when !CYCLE used with !CUMULATIVE

v4.2nd (Beta 15th April 2021)

• Improvements to equation ordering
• Fix bug defining factor classes when factor is past 2500th data field
• Fix for problem using grm1(ID) with !GDENSE
• Fix for problem !NEGBIN was not interpreted correctly
• Allow more characters of 'Level' field to be printed
• Fix for problem where failing to extract initial values in sar2(Row ...)
• Improvements to writing XML

v4.2nc (Beta 20th January 2021)

• Improvements to equation ordering
• Fix to weighted analysis when using grouped variables
• Fix to problem when there are a large number of NRM terms

v4.2nb (Beta 11th December 2020)

• Upgraded Ci reordering for more efficiency
• Improvements to the efficiency for fitting xfa x grm models
• Improvements for grm group processing
• Fix to parsing when using GINDEX
• Fix to parsing when using predict

v4.2nb (Beta 24th September 2020)

• Improved speed for GDENSE
• !LDET as GIV line qualifier.
• Fix to calculation of Gindex standard errors.
• Allow diagonal GRM in .grm file.
• Fix to ARMA model.

v4.2na (Beta 15th April 2020)

• Changes to equation order and other routines for greater efficiency and speed.
• Memory access has been increased; ASReml 4.2 can utilise up to 96GBytes workspace if available.
• The variance function rrk(f) has been added as an alias to xfak(f) with all the specific variances set to 0.
• Tha ability to fit bivariate GL(M)M models has been added.
• Rao-Blackwellized Gibbs sampling is available to faciliate the analysis of large cross classified data sets.
• New design functions: hs(f,k), ref(f,k) and zero(k). See NewFeatures documentation for more details.
• !ARLIMIT to set boundary limit for he magnitude of the AR parameter in spatial models.
• !WVR reports working vari9ables to a .wvr file.
• !WORKSPACE m sepcifies the size of the workspace in GBytes.
• OMP parallel processing has been implemented for some of the major loops.
• Improvements to !PRESENT in predict to speed up calculation.
• More pedigree qualifiers have been added to remove uinneccessary individuals.
• Hierarchical Generalized Linear (Mixed) Models have been implemented and made available for user testing.