[Lme4-commits] r1500 - pkg/lme4Eigen
noreply at r-forge.r-project.org
noreply at r-forge.r-project.org
Mon Jan 9 20:47:29 CET 2012
Author: dmbates
Date: 2012-01-09 20:47:28 +0100 (Mon, 09 Jan 2012)
New Revision: 1500
Modified:
pkg/lme4Eigen/ChangeLog
Log:
Modified: pkg/lme4Eigen/ChangeLog
===================================================================
--- pkg/lme4Eigen/ChangeLog 2012-01-07 23:35:51 UTC (rev 1499)
+++ pkg/lme4Eigen/ChangeLog 2012-01-09 19:47:28 UTC (rev 1500)
@@ -1,5 +1,96 @@
+2012-01-07 Douglas Bates <bates at stat.wisc.edu>
+
+ * man/GHrule.Rd, man/NelderMead-class.Rd, man/NelderMead.Rd,
+ man/golden-class.Rd, man/golden.Rd, man/lmer.Rd, man/mkdevfun.Rd:
+ Documentation updates.
+ * R/lmer.R, R/profile.R, tests/lmer-1.R: Nlmer now working (for nAGQ=1).
+ Updated tests and got profile working again.
+ * R/AllClass.R, R/lmer.R, src/external.cpp, src/optimizer.h: Allowed
+ setting options in the NelderMead optimizer. Switched so that this is
+ the default in glmer.
+ * R/AllClass.R, R/lmer.R, src/external.cpp, src/predModule.cpp: More
+ work on nlmer.
+
+2011-12-31 Douglas Bates <bates at stat.wisc.edu>
+
+ * src/predModule.cpp: Add an explicit copy of the mapped sparse matrix
+ to a sparse matrix in updateL
+
+2011-12-28 Douglas Bates <bates at stat.wisc.edu>
+
+ * R/lmer.R: Change maximum iterations error to a warning.
+ * DESCRIPTION: New version number.
+ * R/lmer.R: Allowed choice of optimizer in glmer at least.
+ * R/AllClass.R: Extended and exported the NelderMead class
+ * src/external.cpp, src/optimizer.cpp, src/optimizer.h: Removed
+ debugging code, added methods to NelderMead to set convergence
+ characteristics and exported same.
+
+2011-12-22 Douglas Bates <bates at stat.wisc.edu>
+
+ * NAMESPACE, R/AllClass.R, src/external.cpp, src/optimizer.cpp,
+ src/optimizer.h: Tuned the Nelder-Mead optimizer a bit and got rid of
+ some of the noise.
+ * R/lmer.R, man/lmer.Rd, src/external.cpp: Incorporated aGQ code in the
+ glmer function itself. Eliminated the doFit argument which is
+ redundant when devFunOnly is also used.
+
+2011-12-21 Douglas Bates <bates at stat.wisc.edu>
+
+ * src/external.cpp: Added the glmerAGQ function and did some cleanup.
+ * src/predModule.cpp: Delete some debugging code.
+ * R/lmer.R: Updates in glmer, aGQ not yet added but soon will be.
+ * R/GHrule.R: Rename the knots on the z scale as 'z', not 'k'
+ * NAMESPACE: Export the golden class and class generator object
+
+2011-12-16 Douglas Bates <bates at stat.wisc.edu>
+
+ * src/respModule.cpp, src/respModule.h: Clean up declarations by
+ declaring a typedef for Eigen::Map<Eigen::VectorXd>
+
+2011-12-15 Douglas Bates <bates at stat.wisc.edu>
+
+ * R/lmer.R: Adopt new version of the merPredD$new function that uses S.
+ * R/GHrule.R: Added an R function, based on SparseGrid:::GQN to evaluate
+ a matrix of nodes, weights and log-density for Gauss-Hermite
+ quadrature against a Gaussian density.
+ * R/AllClass.R, src/optimizer.h: Added a reverse-communication golden
+ search algorithm class
+
+2011-12-12 Douglas Bates <bates at stat.wisc.edu>
+
+ * src/external.cpp: Fat-finger error. I meant to commit this file and
+ not glmFamily.h
+ * src/glmFamily.h, src/optimizer.cpp, src/optimizer.h: Added simple
+ optimizers (Golden search for bounded 1-dimensional and Nelder-Mead
+ for multidimensional).
+
+2011-12-07 Douglas Bates <bates at stat.wisc.edu>
+
+ * R/AllClass.R, R/lmer.R, src/external.cpp, src/predModule.cpp,
+ src/predModule.h: Allow regeneration of pointers in merPredD objects.
+
+2011-12-06 Douglas Bates <bates at stat.wisc.edu>
+
+ * R/AllClass.R, R/lmer.R, src/external.cpp, src/predModule.cpp,
+ src/predModule.h: Add many more fields to the merPredD class and allow
+ for serialize/unserialize. Lower the default tolPwrss as per testing
+ by Ben.
+
+2011-12-05 Douglas Bates <bates at stat.wisc.edu>
+
+ * DESCRIPTION, R/AllClass.R, R/lmer.R, man/lmer.Rd, man/mkdevfun.Rd,
+ man/refitML.Rd, src/external.cpp, src/predModule.cpp,
+ src/predModule.h, src/respModule.cpp, src/respModule.h: Use mapped
+ objects in C++ structures so that serialize/unserialize will work.
+
2011-11-29 Douglas Bates <bates at stat.wisc.edu>
+ * R/profile.R: More fixing of labels. Fix up strip names to
+ reflect sigma, not log(sigma).
+
+2011-11-29 Douglas Bates <bates at stat.wisc.edu>
+
* R/profile.R: Profiling now works for lmer models with variance
components only, including the case of a single fixed-effects
parameter. Labeling of some plots needs to be corrected.
More information about the Lme4-commits
mailing list