Forum: discussion
Monitor Forum | Start New ThreadProject News [ Reply ] By: Vincent Dorie on 2013-09-13 20:19 | [forum:39828] |
News ==== With the update of lme4 to version 1.0 and the change of interals from BLAS/LAPACK to Eigen, blme has undergone a complete rewrite. Major changes: Use of quotations around priors no longer necessary No long automatically rescales priors to data. Rescaling your inputs is probably best policy Minor changes: var.prior argument renamed to resid.prior Can place priors on the square root of the random effect covariance Previously, blme essentially reimplemented lme4 but with priors. The new version uses as much lme4 machinery as possible, with a few downstream impacts. Particularly: Different optimizer means occasionally different results. Also can be a little slower glmms now fit with adaptive Gaussian quadrature for greater accuracy Versioning ========== The main branch of blme will go forward and depend on lme4 version 1.0. A legacy branch will also be provided on R-forge that is compatible with "lme4.0", called blme.0. This info applies until lme4 version 1.0 makes its way to CRAN, at which point using the latest versions should be hassle-free. Finally, a legacy build is available until the CRAN situation has been resolved. It is named "blme" and depends on lme4 for versions less than 1. Installation ============ To install the latest versions: 1) Install lme4 v1.0 1.0) Install appropriate developer tools for your platform. Consult http://cran.r-project.org/faqs.html 1.1) Install devtools package 1.2) Within R, execute require("devtools"); install_github("lme4", user = "lme4", ref = "release") 2) Install blme v0.2 2.1) Within R, execute install.packages("blme", repos = "http://R-Forge.R-project.org") To install legacy versions 1) Install lme4.0 1.1) Within R, execute install.packages("lme4.0", repos = "http://R-Forge.R-project.org", type = "source") 2) Install blme.0 2.1) Within R, execute install.packages("blme.0", repos = "http://R-Forge.R-project.org", type = "source") To install the legacy package 1) Install lme4 version, less than 1 (on CRAN) 2) Install blme legacy 2.1) Checkout blme legacy source by executing from a command line: svn co svn://svn.R-Forge.R-project.org/svnroot/blme/legacy/blme 2.2) Within R, execute install.packages("path/to/blme", repos = NULL, type = "source") |