I have just uploaded the source package for version 0.99875-1 of the lme4 package to CRAN. There are two big additions to this version of the package: - a new function nlmer to fit nonlinear mixed models - a vignette on "Theory and computational methods for mixed models" The vignette is not in its finished form but it is a good portion of the way along to a explaining a uniform approach to linear mixed models, generalized linear mixed models and nonlinear mixed models with crossed and/or nested random effects. If you read this vignette and the other vignette on implementation in the lme4 package then compare them to the code you will see that they describe the lmer2 approach, not the approach that is currently implemented in lmer. Right now I have some implementation issues with generalized linear mixed models and the mcmcsamp function in the lmer2 formulation but I think they are implementation issues and not theoretical roadblocks. Once I resolve those implementation issues I will switch what is now called lmer2 to become lmer. If you have books or papers that record the results of an lmer fit I will be happy to work with you to ensure that the results from the new version are at least as good as the results from the previous version. If you have code that depends on the availability of specific slots in an lmer object you should be aware that your code will break after that switch takes place. I can describe how the lmer2 object represents the model and data but I don't guarantee to fix your code for you if you reached in and grabbed particular slots (as opposed to using extractor functions). Some peculiar results in testing on CRAN in the last few days caused Kurt, Martin and me to look at the function display.mer in package arm. That code makes a lot of assumptions that are going to fail in the near future.
Release 0.99875-1 of the lme4 package uploaded to CRAN
1 message · Douglas Bates