abn can be installed directly from CRAN in the usual way, e.g. from within R using
The abn library makes use of several additional R packages which are not part of CRAN. These are not essential but are strongly recommended as they provide additional functionality.
- INLA which is an R package but hosted separately from CRAN and is easy to install for common platforms (see instructions on INLA website). This package can be used for model fitting.
- Rgraphviz which is part of Bioconductor rather than CRAN. Simple instructions on how to install Bioconductor (which runs within R) are available on the Bioconductor website. This package is used to produce plots of network graphs.
Latest changes and bug fixes
- 18-Jan-2016. Vignette completely restructured with the description of a real data set case study .
- 23-Dec-2014. Updated the function tographviz(), with an additional option to produce undirected graph. Vignette edited according to existing functions.
- 25-Jan-2012. Put in some additional catches in the creation of initial guess estimates in case of design matrix singularity. Auto convert NAs from the buildscorecache() into -.Machine$double.xmax for sending back to GSL routines.
- 04-Jan-2012. Numerous internal changes and increment to version 0.82. Main changes are in how the mlik values in nodes with random effects are estimated. This now uses a one dimensional Brent search as a back up/addition to Nelder-Mead to determine optimal step sizes for estimation of the Hessian.
- 31-Oct-2012. Bug fix. Potential issue in exact searches if mlik scores are given as missing. Internally changed all NaN/NAs to -DBL_MAX to ensure sensible equality comparisons.
- 28-Oct-2012. Bug fix. Marginal estimation in fitabn() could give an error “index out of range” (or similar) if either tail of the density did not go below min.pdf. Now fixed.