Changes in Version 0.2-4
o Additional arguments to lattice::dotplot can now be passed to
plot.(g)lmertree.
o Dedicated plotting function for partitioning growth curves has
been added, see which = "growth" in ?plot.(g)lmertree.
Changes in Version 0.2-3
o Minor bug fixes.
o Experimental function betamertree added, for partitioning
mixed-effects beta regression trees.
Changes in Version 0.2-1
o Bug fix for missing data: Listwise deletion over all predictor,
response and partitioning variables is now applied. Otherwise,
global and local parts of (g)lmertree model could be estimated
based on a different sample size or an error could occur.
o Fixed bug: If dot was used in formula for specifying partitioning
variables, tree fitting would pick up tree structure from previous
iteration. Fixed now.
o Fixed bug: When type = "simple" in call to function plot.(g)lmertree
is specified, further arguments are now passed correctly.
Changes in Version 0.2-0
o Bug fix in predict method: Erroneous allocation of new observations
to tree nodes could occur, which has been fixed now.
o New plotting method for (g)lmertrees. Arguments which, type and fitted
now support wider range of plots. E.g., fitted argument provides
different ways to compute fitted values in terminal panel plots.
E.g., type argument now supports the plotting of fixed-effects
coefficients with standard-error bars.
o Two artificial example datasets added to illustate fitting of constant
fits (MHserviceDemo) and growth-curve models (GrowthCurveDemo) in
terminal nodes.
o New sections added to vignette, illustrating how mixed-effects
regression trees with constant fits and growth curve models in terminal
nodes can be fitted.
Changes in Version 0.1-2
o Functions glmertree() and lmertree() now take offset arguments.
o Argument ranefstart of functions glmertree() and lmertree() can now be
set to TRUE. As a result, the random effects will be estimated
before fitting the tree in the first iteration. This may yield better
results when random effects are expected to be substantial.
o Functions glmertree() and lmertree() now take cluster argument, an optional
vector with cluster IDs to be employed for clustered covariances in the parameter
stability tests.
o Bugs fixed: Argument dfsplit is now passed correctly to tree fitting functions,
additional arguments are now passed correctly to (g)lmer(). .
o Arguments "lmer.control" in lmertree() and "glmer.control" in glmertree()
are now actually passed to lmer() and glmer() internally.
Changes in Version 0.1-1
o First CRAN release of the "glmertree" package for fitting generalized linear
mixed-effects model trees in R. For an introduction to the underlying methods see
Fokkema, Smits, Zeileis, Hothorn, Kelderman (2015). Detecting
Treatment-Subgroup Interactions in Clustered Data with Generalized
Linear Mixed-Effects Model Trees. Working Paper 2015-10. Working Papers
in Economics and Statistics, Research Platform Empirical and Experimental
Economics, Universitaet Innsbruck.
URL http://EconPapers.RePEc.org/RePEc:inn:wpaper:2015-10
The package is under development on R-Forge at
https://R-Forge.R-project.org/projects/partykit/