ASReml-R

This version offers a more unified framework and extended functionality for LMM analysis, particularly for large and complex data sets. New features with Version 4 include:

• new and improved licensing
• a simplified and more meaningful syntax for model specification:
  • rcov becomes residual and direct sum structures for residual models with data partitioned into sections to which separate variance structures are applied are now logically specified using a dsum() function
  • rarely used options are now listed externally to the model call in a new asreml.options() function
• a more unified framework for related arguments in the call to asreml()and the output object , e.g.
  • na.action() for dealing with missing values in both the response and explanatory variables
  • a formulae component of the ASReml-R object which lists the fixed, random, sparse and residual model formulae
• improved updates in factor analytic models and a reduced rank rr() variant of the fa() variance model function
• a simpler, more consistent specification of known variance models (including relationship matrices) through the vm() function, which also caters for known singular matrices
• computationally efficient fitting of random regression models when there are more variables than observations – motivated by the use of SNP marker data to explain genotypes
• more informative warning and error messages
• improved graphics powered by ggplot2.

More advanced functionality new with Version 4 includes:

• introduction of the own() variance model to allow the specification of a user-defined variance structure
• extensions to generalized linear models including threshold models and bivariate models with one variate having a normal distribution and the other variate distributed from an exponential family distribution
• generating design matrices to allow use of derived model terms and functions; design argument to asreml.options()
• functions to generate factors that either combine levels of a factor or use a subset of levels to allow easier prediction of models; combine, prune, gpf(), sbs().

Also included:

• computing functions of variance components and their approximate standard errors; vpredict()
• calculating information criteria including AIC and BIC.

This is just a selection from the full set of new features available with ASReml-R Version 4. A detailed account of the new functionality is presented in a navigation guide which guides existing users in transitioning from Version 3 to Version 4.