lme4
statusVersion 2013-05-08 17:03:31
Almost on the list is "how can we release the damn thing?"
lme4
repository, etc.)debug
branch, with build from source (even after a clean Github checkout), with example(glmer)
(lmer
is OK). BMB can't reproduce any more, on either Ubuntu or MacOS.install_github
, git clone and install from source, install from tarball, various versions of upstream packages (CRAN vs R-forge), etc. (This is not going to happen any time soon!)otherwise, as far as we know, this is resolved -- probably want to put out some sort of vague warning / "let us know if you see anything fishy" to r-sig-mixed-models
GLMM deviance function / PWRSS issues
as far as BMB recalls, brute-force resetting fixes every problem with possible exception of some very non-canonical links (which didn't work before anyway! and therefore are not release-critical)
downstream packages: general
finish getting pkgtests
working, follow up on issues with downstream developers (MM said long ago that he would be willing to help with this)
downstream packages: blme
blme
to look for lme4.0
rather than lme4
(i.e. search and replace lme4.0
-> lme4
)highest priority: priors on variance-covariance matrices, and glmer
with family="gaussian"
to get (slow) priors on $\beta$
PAPERS
glmer
/blme
issueslme4
unstable: ideally with working (non-fragile) GLMMs, but possibly as a fallback position with only LMMs fully supporteddevComp
to include a native-R implementation?lme4-extras
vignette (still on r-forge, not copied yet to github)Preparatory to dumping things in DB/MM's lap and whining to them to fix RcppEigen so that it can handle them ...
lme4
to show that it fails at the initial Cholesky decomposition of "B" (i.e. constructing RZX
) (2) written out and checked native-R construction of all of the bits leading up to B
(3) shown that Matrix::Cholesky
gives us a sensible answer for the decomposition. We haven't (4) instrumented lme4.0
to show it does the same thing (5) written a base::chol
example to show that it gets the same answer as Matrix::Cholesky (6) run
RcppEigen::solveInPlacein a minimal code file outside of
lme4(7) do a thorough writeup of what's going on, including a
.RDatafile that has all the pieces saved in it (i.e. original data,
Z,
X,
B,
RZX`, ...) [#4 and #6 are not necessary]nlmer
example? (algal example)We found this useful in discussing lme4
issues
lme4
column-structured-branch stuff, etc.nlminb
, bobyqa
, etc. etc.. (e.g. Koller boundary-sticking problems)Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.