Rbayz

Bayesian mixed models, shrinkage, sparse and interaction kernel regression

Build StatusCoverage Status coveralls

Sections in this page - About R/bayz - Obtaining R/bayz - A very short tour - Example - Summarizing and using output

About R/bayz

R/bayz is in the basis a mixed model packages, but it can do much more than what other common R mixed model packages offer. R/bayz has wide support to insert covariance / similarity / relationship structures on random effects, either through a supplied “kernel”, with modelled structures such as spatial, unstructured and factor analytic covariance structures, and variaous approaches to model heterogeneous variances. For interactions of multiple random effects, R/bayz can build their covariance structure as a Kronecker product, and these Kronecker products are handled efficiently by avoiding to explicitly build the product matrix. A common application of this in plant breeding is to model Cultivar:Environment interactions with a Kronecker product of a genomic and a enviromic similarity matrix. R/bayz can go beyond interactions and Kronecker products of two terms, for instance, allowing to add a Trait dimension as well, to fit large multi-trait models using a Factor Analytic approach. R/bayz also supports using (multiple) large sets of predictors that can be modelled with different shrinkage regression options. Lastly, being Bayesian, R/bayz goes beyond standard mixed model capabilities by allowing different shrinkage and sparse shrinkage distributions on effects, as well as on kernels.

The R help ?bayz contains basic help information; full details are available on <ljanss.github.io/Rbayz>.

Obtaining R/bayz

R/bayz us currently not an officially (cran) released R package. It can be downloaded and istalled with the options shown below. Not being on cran implies that dependencies need to be installed manually, which is also shown below.

1. Precompiled binary versions (Windows, Mac)

You can try your luck if you can install some of the following precompiled binary packages for Windows and MacOS using the below commands in the R terminal. If this fails, go to option 2 to install source from Github.

Installing package dependencies:

install.packages("Rcpp")
install.packages("nlme")
install.packages("coda")

For Windows (compiled on Win11 64-bit)

install.packages("https://ljanss.github.io/Rbayz/Rbayz_0.9.0.zip",repos=NULL,type="win.binary")

For MacOS (compiled on Mac Silicon)

install.packages("https://ljanss.github.io/Rbayz/Rbayz_0.9.0.tgz",repos=NULL,type="mac.binary")

2. Install source from Github repository (all systems including linux)

This requires a development environment in R, which needs Rtools on windows, or “command line tools” on MacOS, and the devtools package. On linux the development tools may often be pre-installed. The below commands run in the R terminal.

Installing devtools and package dependencies:

install.packages("devtools")
install.packages("Rcpp")
install.packages("nlme")
install.packages("coda")

Download and install/compile Rbayz using:

library(devtools)
devtools::install_github("ljanss/Rbayz")

A very short tour of R/bayz

R/bayz has a single main function bayz() that accepts model formulas in an extended R-formula syntax where all explanatory (right-hand-side) variables are wrapped by a “function-like” to specify how to fit it in the model. This may look like Yield ~ fx(Year) + rn(Variety) to fit Yield with Year as a fixed factor and Variety as a random factor (fx() and rn() imply the variable used should be (converted to) factor). For random effects, variance structures can be added with a V= option inside the rn(), for instance with a relationship / similarity matrix rn(Variety,V=Gmat), which implies the variance-covariance structure Gmatσg2.

Interactions between fators are specified using the colon, such as fx(Year:Location) for fixed effects of Year-Location interactions. R/bayz does not support automatic expansion with main effects, and main effects, if desired, should be explicitly added in the model. For interactions in random effects, the variance specification is expanded to a product of terms (interpreted as Kronecker product), written as rn(Variety:Environment, V=Gmat*Emat), implying the variance-covariance structure for the interaction effects as Gmat* ⊗ *Emat**σg**e2. Using interactions and Kronecker products involving estimated unstructured covariance matrices allows to fit multi-trait models by supplying data in a “melted” form and specifying interactions with Trait. Interactions and the matching variance structures can be specified to any order. R/bayz can additionally fit large sets of predictors using the rr() (ridge or random regression) model function, regular fixed (nested) regressions using rg(), and random slope models using rs().

R/bayz is used as many other R model functions by capturing the output object, and running methods on that output like summary(), plot(), various methods to extract part of the estimates, and a contrast and predict method. The summary() gives MCMC convergence diagnostics and Highest Posterior Density regions for selected “traced” parameters, while plot() produces trace plots and density plots.

Check the following manual pages for additional information on: